We analyze Elman-type recurrent neural networks (RNNs) and their training in the mean-field regime. Specifically, we show convergence of gradient descent training dynamics of the RNN to the corresponding mean-field formulation in the large width limit. We also show that the fixed points of the limiting infinite-width dynamics are globally optimal, under some assumptions on the initialization of the weights. Our results establish optimality for feature-learning with wide RNNs in the mean-field regime.

BACKGROUND: Common data models (CDMs) are essential tools for data harmonization, which can lead to significant improvements in the health domain. CDMs unite data from disparate sources and ease collaborations across institutions, resulting in the generation of large standardized data repositories across different entities. An overview of existing CDMs and methods used to develop these data sets may assist in the development process of future models for the health domain, such as for decision support systems.OBJECTIVE: This scoping review investigates methods used in the development of CDMs for health data. We aim to provide a broad overview of approaches and guidelines that are used in the development of CDMs (ie, common data elements or common data sets) for different health domains on an international level.METHODS: This scoping review followed the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) checklist. We conducted the literature search in prominent databases, namely, PubMed, Web of Science, Science Direct, and Scopus, starting from January 2000 until March 2022. We identified and screened 1309 articles. The included articles were evaluated based on the type of adopted method, which was used in the conception, users' needs collection, implementation, and evaluation phases of CDMs, and whether stakeholders (such as medical experts, patients' representatives, and IT staff) were involved during the process. Moreover, the models were grouped into iterative or linear types based on the imperativeness of the stages during development.RESULTS: We finally identified 59 articles that fit our eligibility criteria. Of these articles, 45 specifically focused on common medical conditions, 10 focused on rare medical conditions, and the remaining 4 focused on both conditions. The development process usually involved stakeholders but in different ways (eg, working group meetings, Delphi approaches, interviews, and questionnaires). Twenty-two models followed an iterative process.CONCLUSIONS: The included articles showed the diversity of methods used to develop a CDM in different domains of health. We highlight the need for more specialized CDM development methods in the health domain and propose a suggestive development process that might ease the development of CDMs in the health domain in the future.

We present Spacerini, a tool that integrates the Pyserini toolkit for reproducible information retrieval research with Hugging Face to enable the seamless construction and deployment of interactive search engines. Spacerini makes state-of-the-art sparse and dense retrieval models more accessible to non-IR practitioners while minimizing deployment effort. This is useful for NLP researchers who want to better understand and validate their research by performing qualitative analyses of training corpora, for IR researchers who want to demonstrate new retrieval models integrated into the growing Pyserini ecosystem, and for third parties reproducing the work of other researchers. Spacerini is open source and includes utilities for loading, preprocessing, indexing, and deploying search engines locally and remotely. We demonstrate a portfolio of 13 search engines created with Spacerini for different use cases.

Will protein structure search tools like AlphaFold replace protein sequence search with BLAST? We discuss the promises, using structure search for remote homology detection, and why protein BLAST, as the leading sequence search tool, should strive to incorporate structural information

BACKGROUND: Integrating multi-omics data is fast becoming a powerful approach for predicting disease progression and treatment outcomes. In light of that, we introduce a modified version of the NetRank algorithm, a network-based algorithm for biomarker discovery that incorporates the protein associations, co-expressions, and functions with its phenotypic association to differentiate different types of cancer. NetRank is introduced here as a robust feature selection method for biomarker selection in cancer prediction. We assess the robustness and suitability of the RNA gene expression data through scanning genomic data for 19 cancer types with more than 3000 patients from The Cancer Genome Atlas (TCGA).RESULTS: The results of evaluating different cancer type profiles from the TCGA data demonstrate the strength of our approach to identifying interpretable biomarker signatures for cancer outcome prediction. NetRank's biomarkers segregate most cancer types with an area under the curve (AUC) above 90% using compact signatures.CONCLUSION: In this paper we provide a fast and efficient implementation of NetRank, with a case study from The Cancer Genome Atlas, to assess the performance. We incorporated complete functionality for pre and post-processing for RNA-seq gene expression data with functions for building protein-protein interaction networks. The source code of NetRank is freely available (at github.com/Alfatlawi/Omics-NetRank) with an installable R library. We also deliver a comprehensive practical user manual with examples and data attached to this paper.

Rad52 is a highly conserved eukaryotic protein that can mediate the annealing of complementary DNA strands to initiate homologous recombination for the repair of double-strand breaks. Suspicions that at least some prokaryotic single-strand annealing proteins (SSAPs) are related to Rad52 have been discussed for more than two decades. Two recent cryo-EM structures, now put the issue beyond doubt.

The BigCode project is an open-scientific collaboration working on the responsible development of large language models for code. This tech report describes the progress of the collaboration until December 2022, outlining the current state of the Personally Identifiable Information (PII) redaction pipeline, the experiments conducted to de-risk the model architecture, and the experiments investigating better preprocessing methods for the training data. We train 1.1B parameter models on the Java, JavaScript, and Python subsets of The Stack and evaluate them on the MultiPL-E text-to-code benchmark. We find that more aggressive filtering of near-duplicates can further boost performance and, surprisingly, that selecting files from repositories with 5+ GitHub stars deteriorates performance significantly. Our best model outperforms previous open-source multilingual code generation models (InCoder-6.7B and CodeGen-Multi-2.7B) in both left-to-right generation and infilling on the Java, JavaScript, and Python portions of MultiPL-E, despite being a substantially smaller model. All models are released under an OpenRAIL license at https://hf.co/bigcode.

The question of how a given knowledge base can be modified such that certain unwanted consequences are removed has been investigated in the area of knowledge engineering under the name of repair and in the area of belief change under the name of contraction. Whereas in the former area the emphasis was more on designing and implementing concrete repair algorithms, the latter area concentrated on characterizing classes of contraction operations by certain postulates they satisfy. In the classical setting, repairs and contractions are subsets of the knowledge base that no longer have the unwanted consequence. This makes these approaches syntax-dependent and may result in removal of more consequences than necessary. To alleviate this problem, gentle repairs and pseudo-constractions have been introduced in the respective research areas, and their connections have been investigated in recent work. Optimal repairs preserve a maximal amount of consequences, but they may not always exist. We show that, if they exist, then they can be obtained by certain pseudo-contraction operations, and thus they comply with the postulates that these operations satisfy. Conversely, under certain conditions, pseudo-contractions are guaranteed to produce optimal repairs.

The question of how a given knowledge base can be modified such that certain unwanted consequences are removed has been investigated in the area of ontology engineering under the name of repair and in the area of belief change under the name of contraction. Whereas in the former area the emphasis was more on designing and implementing concrete repair algorithms, the latter area concentrated on characterizing classes of contraction operations by certain postulates they satisfy. In the classical setting, repairs and contractions are subsets of the knowledge base that no longer have the unwanted consequence. This makes these approaches syntax-dependent and may result in removal of more consequences than necessary. To alleviate this problem, gentle repairs and pseudo-constractions have been introduced in the respective research areas, and their connections have been investigated in recent work. Optimal repairs preserve a maximal amount of consequences, but they may not always exist. We show that, if they exist, then they can be obtained by certain pseudo-contraction operations, and thus they comply with the postulates that these operations satisfy. Conversely, under certain conditions, pseudo-contractions are guaranteed to produce optimal repairs. Recently, contraction operations have also been defined for concepts rather than for whole knowledge bases. We show that there is again a close connection between such operations and optimal repairs of a restricted form of knowledge bases.

Der Einsatz von Learning Analytics etabliert sich zunehmend an Hochschulen. Im Rahmen eines vom BMBWF kofinanzierten Projekts wurde sowohl auf Ebene der Lehrveranstaltung als auch auf Ebene des Studiums angesetzt. Tools und Qualifizierungsmaßnahmen wurden entwickelt, die Studierende und ihre Interessen ins Zentrum stellen. Das Projekt verfolgt somit einen studierendenzentrierten Ansatz, der Autonomie und Freiwilligkeit der Teilnahme in den Vordergrund rückt. An den drei beteiligten Universitäten wurden Dashboards, wie der Learner's Corner sowie Begleitmaßnahmen (Mentoring, Tutorials, …) entwickelt und empirisch überprüft. Außerdem wurden ethische und datenschutzrechtliche Richtlinien im Rahmen eines interdisziplinär erarbeiteten Kriterienkatalogs entwickelt. Im Beitrag wird der Einsatz von Learning-Analytics-Tools auf unterschiedlichen Ebenen (Moodle-Plug-ins, Anwendung zur Visualisierung des Studienfortschritts) anhand von aus dem Projekt ausgewählten Schwerpunkten exemplarisch veranschaulicht und einzelne empirische Ergebnisse werden vorgestellt.

The paper deals with the topic of realizability in abstract argumentation. More precisely, we consider the most basic kind of labellings, so-called conflict-free labellings. The understanding of these labellings is essential as any mature labelling-based semantics selects its output among these labellings. We show how to decide whether a given set of labellings is the conflict-free outcome of a certain Dungean framework. To this end we introduce several new criteria like L-tightness, reject-witness and reject-compositionality. These properties play a decisive role in the central characterization theorem. Moreover, we present a construction method, showing how to realize a conflict-free realizable labelling-set. Finally, we study the representational freedom in case of such labellings. This leads to an uniqueness result for a certain sub-class and a surprising connection to strong equivalence in case of extension-based semantics.

We address the issue of forgetting in assumption-based argumentation (ABA). Forgetting is driven by the goal to remove certain elements from a knowledge base, while preserving the structure of its models as well as possible. We introduce several forgetting operators tailored to accomplish the removal of different pieces of the ABA knowledge base---assumptions, contraries, and atoms---formalizing a diverse selection of perspectives on this issue. We examine the quality of our operators by studying their compliance with suitable desiderata we propose. Thereby, we investigate the impact of the operators on the syntax of the given ABA knowledge base, its semantics, but also the instantiated argumentation framework; thus bridging recent forgetting studies on non-monotonic formalisms including argumentation theory.

INTRODUCTION: Obese rodents e.g., the leptin-deficient (ob/ob) mouse exhibit remarkable behavioral changes and are therefore ideal models for evaluating mental disorders resulting from obesity. In doing so, female as well as male ob/ob mice at 8, 24, and 40 weeks of age underwent two common behavioral tests, namely the Open Field test and Elevated Plus Maze, to investigate behavioral alteration in a sex- and age dependent manner. The accuracy of these tests is often dependent on the observer that can subjectively influence the data.METHODS: To avoid this bias, mice were tracked with a video system. Video files were further analyzed by the compared use of two software, namely EthoVision (EV) and DeepLabCut (DLC). In DLC a Deep Learning application forms the basis for using artificial intelligence in behavioral research in the future, also with regard to the reduction of animal numbers.RESULTS: After no sex and partly also no age-related differences were found, comparison revealed that both software lead to almost identical results and are therefore similar in their basic outcomes, especially in the determination of velocity and total distance movement. Moreover, we observed additional benefits of DLC compared to EV as it enabled the interpretation of more complex behavior, such as rearing and leaning, in an automated manner.DISCUSSION: Based on the comparable results from both software, our study can serve as a starting point for investigating behavioral alterations in preclinical studies of obesity by using DLC to optimize and probably to predict behavioral observations in the future.

We formulate an ergodic theory for the (almost sure) limit PE˜co of a sequence (PEnco) of successive dynamic imprecise probability kinematics (DIPK, introduced in [10]) updates of a set PE0co representing the initial beliefs of an agent. As a consequence, we formulate a strong law of large numbers.

Water resources management challenges are multiple and complex, and human force, as the main driver of environmental change, has been increasing the need for tailored (and faster) responses to them over the past decades. Despite our increasing technical knowledge on how to tackle these issues, which are mainly related to water quantity, quality and access, unprecedented change in climate and landscapes will require a better understanding of the interactions between water and society. This thesis is concerned with the challenging task of applying machine learning techniques to extract knowledge from hydrological, socioeconomic and climate data and tackle some of the water management issues associated with water quantity and quality. Specifically, it addresses (i) the drivers of water demand in different temporal and spatial scales; (ii) the implications of price-based demand-side measures and how media coverage and public interest on extreme events can affect consumption habits; (iii) the long-term water availability and supply under climate variability, and (iv) some of the effects of environmental change on water quality. We learn that climate variability and change might affect not only hydrological responses but also consumption habits and water supply expansion strategies. Also, we make valuable findings on the drivers of water demand and quality, which can support utilities in their long-term planning.

Water allocation during droughts is a challenge for policymakers, often addressed through participatory approaches. The implications of this governance mode are understudied as long-term records of the decision-making processes are often unavailable. We use natural language processing (NLP) and network analysis to extract information on water allocation decisions and climate-related issues from meeting minutes of river basin stakeholders. To test this approach, we considered the minutes of 1100 meetings held between 1997 and 2021 in the twelve basin committees of Ceará, Brazil. This region has a long history of droughts, which have strongly influenced water policies and politics. The river basin committee is currently composed of representatives of governmental and non-governmental institutions and deliberates on the water management process. To identify conflicts and relevant issues discussed …

In an ever-evolving digital world with complex algorithms like machine learning, we need new strategies for more flexibility to cope with the ever-changing environment. For this, runtime scalability and runtime adaptability for low-power and highly efficient hardware is a promising solution. By combining the runtime reconfiguration of FPGAs with the efficient communication of Networks-on-Chip (NoC), we are able to implement a highly scalable, high-performance, and energy-efficient computing architecture that fixed-function units and specialized static accelerators lack. In this work, we introduce a RunTime Adaptable and Scalable System for NoC-based architectures called RTASS. The hardware architecture includes a master subsystem, a network adapter, and an NoC subsystem with parametrizable routers and several various routing algorithms. Furthermore, RTASS provides a software architecture that includes advanced drivers for runtime management. The key benefit of RTASS is the ability to dynamically adjust the number of routers within the NoC at runtime based on the current application's requirements. That allows the system to support both homogeneous and inhomogeneous types of processing elements as well as regular and irregular shapes. The development of this runtime scalable and flexible architecture will establish the foundation for future highly adaptable applications such as machine learning and computer vision in the embedded computing field. We implemented and evaluated the proposed work with the Xilinx Zynq-7000 FPGA, with the possibility of porting it to other FPGAs that support runtime reconfiguration.

Large-scale information on several vegetation properties (‘plant traits’) is critical to assess ecosystem functioning, functional diversity and their role in the Earth system. Hyperspectral remote sensing of plant canopies offers a key tool to map multiple plant traits. However, we are still lacking generalized methods to translate hyperspectral reflectance into a suite of relevant plant traits across biomes, land cover and sensor types. The absence of globally representative data sets and the gap between the available reflectance data with corresponding in-situ measurements have hampered such approaches. In recent years, the scientific community acquired multiple data sets encompassing canopy hyperspectral reflectance and plant traits from different plant types and sensors. To combine these heterogeneous data sets, we propose three multi-trait modeling approaches based on Convolutional Neural Networks (CNNs …

Our understanding of the Earth's functional biodiversity and its imprint on ecosystem functioning, structure and resilience is still incomplete. Large-scale information on vegetation properties ('plant traits') is critical to assess functional diversity and its role in the Earth system. Such parameters are constantly changing due to variations in environmental conditions which makes extensive in-situ measurements in the field not logistically feasible. The advent of the upcoming space-borne hyperspectral missions will facilitate to map these properties. However, we are still lacking efficient and accurate methods to translate hyperspectral reflectance into large scale information on plant traits across biomes, land cover and sensor types. Yet, the absence of globally representative data sets on reflectance data and the corresponding in-situ measurements represents a bottleneck to develop empirical models for estimating plant …

Over the past decade, researchers have developed several new methods to create microscopic machines that move in a liquid environment by absorbing energy. This self-propelled motion of the so-called microswimmers not only mimics a fundamental element of living systems, but also follows the vision put forward by the 1966 movie Fanstatic Voyage and many others since. The idea is that machines can be miniaturized so that they can be injected into the human body for use against injury and disease. While the idea at the time still involved miniaturizing humans to control the machines, today's research on machine learning and artificial intelligence, which also roots back to the 1960s, assumes that such machines could be intelligent themselves. While this is still a dream, research on the active matter in conjunction with machine learning goes far beyond the perspectives of healthcare but considers synthetic micromachines as novel building blocks of adaptive self-organized or even intelligent materials for information processing itself. The following chapter presents some basic elements of how machine learning can be used to enhance the motion of small micromachines or active particles. In particular, it focuses on reinforcement learning as an exploratory learning approach that is demonstrated in variants to control single or multiple microswimmers in the experiment.

AbstractSETD1A is the histone 3 lysine 4 (H3K4) methyltransferase central to the mammalian version of the highly conserved eight subunit Set1 complex (Set1C) that apparently conveys H3K4 trimethylation (H3K4me3) onto all active Pol II promoters. Accordingly, mouse embryonic stem cells (ESCs) die when SETD1A is removed. We report that death is accompanied by loss of expression of DNA repair genes and accumulating DNA damage. BOD1L and BOD1 are homologs of the yeast Set1C subunit, Shg1, and subunits of the mammalian SETD1A and B complexes. We show that the Shg1 homology region binds to a highly conserved central $alpha$-helix in SETD1A and B. Like mutagenesis ofShg1in yeast, conditional mutagenesis ofBod1lin ESCs promoted increased H3K4 di- and tri-methylation but also, like loss of SETD1A, loss of expression of DNA repair genes, increased DNA damage and cell death. In contrast to similar losses of gene expression, the converse changes in H3K4 methylation implies that H3K4 methylation is not essential for expression of the DNA repair network genes. Because BOD1L becomes highly phosphorylated after DNA damage and acts to protect damaged replication forks, the SETD1A complex and BOD1L in particular are key nodes for the DNA damage repair network.

Concrete domains have been introduced in Description Logic (DL) to enable reference to concrete objects (such as numbers) and predefined predicates on these objects (such as numerical comparisons) when defining concepts. The primary research goal in this context was to find restrictions on the concrete domain such that its integration into certain DLs preserves decidability or tractability. In this paper, we investigate the abstract expressive power of logics extended with concrete domains, namely whichclasses of first-order interpretations can be expressed using these logics. In the first part of the paper, we show that, under natural conditions on the concrete domain D (which also play a role for decidability), extensions of first-order logic (FOL) or 풜ℒ풞 with D share important formal properties with FOL, such as the compactness and the Löwenheim-Skolem property. Nevertheless, their abstract expressive power need not be contained in that of FOL. In the second part of the paper, we investigate whether finitely bounded homogeneous structures, which preserve decidability if employed as concrete domains, can be used to express certain universal first-order sentences, which then could be added to DL knowledge bases without destroying decidability. We show that this requires rather strong conditions on said sentences or an extended scheme for integrating the concrete domain that leads to undecidability.

In this paper we report the benchmarking results of four algorithms on the Strict Box-Constraint Optimization Studies (SBOX-COST) benchmarking suite which, as the name states, employs strict box constraints. These results are compared to the benchmarking results on the Black-box Optimization Benchmarking (BBOB) suite which serves as basis for the SBOX-COST suite, but is less restrivtive with respect to box constraints. SBOX-COST enforces its box constraints by returning an invalid value (∞) whenever a point outside of the bounds is evaluated. We use the following four algorithms from the Nevergrad Toolbox: Estimation of Multivariate Normal Algorithm (EMNA), Differential Evolution (DE), Constrained Optimization BY Linear Approximation (Cobyla) and Particle Swarm Optimization (PSO). All algorithms are employed without parameter tuning. Generally, all algorithms perform quite similiar on both suites but we notice a slight advantage for the algorithms that are allowed to evaluate the functions outside of the bounds. The effect of the advantage varies very much with function ID and in some cases, an algorithm's performance on a problem from the SBOX-COST suite is even better than on the corresponding problem from the BBOB suite.

Open Artificial Intelligence (Open source AI) collaboratives offer alternative pathways for how AI can be developed beyond well-resourced technology companies and who can be a part of the process. To understand how and why they work and what additionality they bring to the landscape, we focus on three such communities, each focused on a different kind of activity around AI: building models (BigScience workshop), tools and ways of working (The Turing Way), and ecosystems (Mozilla Festival's Building Trustworthy AI Working Group). First, we document the community structures that facilitate these distributed, volunteer-led teams, comparing the collaboration styles that drive each group towards their specific goals. Through interviews with community leaders, we map user journeys for how members discover, join, contribute, and participate. Ultimately, this paper aims to highlight the diversity of AI work and workers that have come forth through these collaborations and how they offer a broader practice of openness to the AI space.

A central element of any effective teaching and learning strategy is assessment. Closely related to evaluations of learning results is the constant challenge on the part of higher education teachers to develop assessment strategies aligned to the intended learning outcomes. The Corona-induced wave of digitation in higher education in spring 2020 opened up new forms of assessment and assessment approaches which are closely related to changing teaching and learning methods. The aim of this study is to explore how digital examination differs from those conducted in person. In order to make distinctions in the areas of examination formats, designs and strategies for examination implementation, guided interviews with 12 university lecturers at the University of Graz were conducted. Additionally, 14 examination forms from face-to-face and online teaching were analyzed. The results show that examination strategies as well as examination formats are strongly dependent on the size of the course and need to be adapted to the given framework conditions. Online examinations, comprehension and transfer questions as well as open-book examinations are increasingly used to counteract the reduced control in the online setting.

Der Einsatz von Learning Analytics etabliert sich zunehmend an Hochschulen. Im Rahmen eines vom BMBWF kofinanzierten Projekts wurde sowohl auf Ebene der Lehrveranstaltung als auch auf Ebene des Studiums angesetzt. Tools und Qualifizierungsmaßnahmen wurden entwickelt, die Studierende und ihre Interessen ins Zentrum stellen. Das Projekt verfolgt somit einen studierendenzentrierten Ansatz, der Autonomie und Freiwilligkeit der Teilnahme in den Vordergrund rückt. An den drei beteiligten Universitäten wurden Dashboards, wie der Learnertextquoterights Corner sowie Begleitmaßnahmen (Mentoring, Tutorials, …) entwickelt und empirisch überprüft. Außerdem wurden ethische und datenschutzrechtliche Richtlinien im Rahmen eines interdisziplinär erarbeiteten Kriterienkatalogs entwickelt. Im Beitrag wird der Einsatz von Learning-Analytics-Tools auf unterschiedlichen Ebenen (Moodle-Plug-ins, Anwendung zur Visualisierung des Studienfortschritts) anhand von aus dem Projekt ausgewählten Schwerpunkten exemplarisch veranschaulicht und einzelne empirische Ergebnisse werden vorgestellt.

BACKGROUND: The overall survival of patients with advanced and refractory oesophageal squamous cell carcinoma, mostly aged 65 years and older, is poor. Treatment with PD-1 antibodies showed improved progression-free survival and overall survival. We assessed the safety and efficacy of combined nivolumab and ipilimumab therapy in this population. METHODS: This multicentre, open-label, phase 2 trial done in 32 sites in Germany included patients aged 65 years and older with oesophageal squamous cell carcinoma and disease progression or recurrence following first-line therapy. Patients were treated with nivolumab (240 mg fixed dose once every 2 weeks, intravenously) in the safety run-in phase and continued with nivolumab and ipilimumab (nivolumab 240 mg fixed dose once every 2 weeks and ipilimumab 1 mg/kg once every 6 weeks, intravenously). The primary endpoint was overall survival, which was compared with a historical cohort receiving standard chemotherapy in the intention-to-treat population. This study is registered with ClinicalTrials.gov, NCT03416244. FINDINGS: Between March 2, 2018, and Aug 20, 2020, we screened 75 patients with advanced oesophageal squamous cell carcinoma. We enrolled 66 patients (50 [76%] men and 16 [24%] women; median age 70·5 years [IQR 67·0-76·0]), 44 (67%) of whom received combined nivolumab and ipilimumab therapy and 22 (33%) received nivolumab alone. Median overall survival time at the prespecified data cutoff was 7·2 months (95% CI 5·7-12·4) and significantly higher than in a historical cohort receiving standard chemotherapy (p=0·0063). The most common treatment-related adverse events were fatigue (12 [29%] of 42), nausea (11 [26%]), and diarrhoea (ten [24%]). Grade 3-5 treatment-related adverse events occurred in 13 (20%) of 66 patients. Treatment-related death occurred in one patient with bronchiolitis obliterans while on nivolumab and ipilimumab treatment. INTERPRETATION: Patients aged at least 65 years, with advanced oesophageal squamous cell carcinoma might benefit from combined nivolumab and ipilimumab therapy in second-line treatment. FUNDING: Bristol Myers Squibb.

Mixed Reality (MR) popularizes numerous situated applications where virtual content is spatially integrated into our physical environment. However, we only know little about what properties of an environment influence the way how people place digital content and perceive the resulting layout. We thus conducted a preliminary study (N = 8) examining how physical surfaces affect organizing virtual content like documents or charts, focusing on user perception and experience. We found, among others, that the situated layout of virtual content in its environment can be characterized by the level of spatial as well as semantic coupling. Consequently, we propose a two-dimensional design space to establish the vocabularies and detail their parameters for content organization. With our work, we aim to facilitate communication between designers or researchers, inform general MR interface design, and provide a first step towards future MR workspaces empowered by blending digital content and its real-world context.

In this paper, we discuss the benefits and challenges of shared tasks as a teaching method. A shared task is a scientific event and a friendly competition to solve a research problem, the task. In terms of linking research and teaching, shared-task-based tutorials fulfill several faculty desires: they leverage students' interdisciplinary and heterogeneous skills, foster teamwork, and engage them in creative work that has the potential to produce original research contributions. Based on ten information retrieval (IR) courses at two universities since 2019 with shared tasks as tutorials, we derive a domain-neutral process model to capture the respective tutorial structure. Meanwhile, our teaching method has been adopted by other universities in IR courses, but also in other areas of AI such as natural language processing and robotics.

When asked, large language models (LLMs) like ChatGPT claim that they can assist with relevance judgments but it is not clear whether automated judgments can reliably be used in evaluations of retrieval systems. In this perspectives paper, we discuss possible ways for LLMs to support relevance judgments along with concerns and issues that arise. We devise a human--machine collaboration spectrum that allows to categorize different relevance judgment strategies, based on how much humans rely on machines. For the extreme point of ``fully automated judgments'', we further include a pilot experiment on whether LLM-based relevance judgments correlate with judgments from trained human assessors. We conclude the paper by providing opposing perspectives for and against the use of~LLMs for automatic relevance judgments, and a compromise perspective, informed by our analyses of the literature, our preliminary experimental evidence, and our experience as IR researchers.

Answering temporal CQs over temporalized Description Logic knowledge bases (TKB) is a main technique to realize ontology-based situation recognition. In case the collected data in such a knowledge base is inaccurate, important query answers can be missed. In this paper we introduce the TKB Alignment problem, which computes a variant of the TKB that minimally changes the TKB, but entails the given temporal CQ and is in that sense (cost-) optimal. We investigate this problem for ALC TKBs and conjunctive queries with LTL operators and devise a solution technique to compute (cost-optimal) alignments of TKBs that extends techniques for the alignment problem for propositional LTL over finite traces.

Drought is a natural hazard with complex socioeconomic impacts and influences on human experiences. Preparedness is the only way a society can mitigate drought impacts but integrating participatory decision-making with hydrological modeling into a more comprehensive planning process is still a challenge. Here, we present a step-by-step methodology to guide the implementation of a participatory drought preparedness plan (DPP), specially designed for hydrosystems and cities scales. We highlight strategies to engage local stakeholders in constructing such plans and build trust in the process. We propose two types of drought preparedness plans: (1) Socio-technical–built only from the tacit knowledge of the system operators, which needs only two days to be ready; and (2) Socio-technical with modeling-intensive simulation—a more robust methodology that adds hydrologic and hydraulic modeling to the …

In recent years, High Performance Computing (HPC) has become increasingly important for many industries and research areas besides textquoteleftclassictextquoteright applications. As new domains emerge, applications, implementations and frameworks become more diverse. Generic performance analysis tools often cannot keep up with the development speed of new approaches for workload distribution, offloading, and communication. Some of the new approaches employ their own performance monitoring, which is difficult to integrate into generic tools designed for traditional HPC. Performance measurements often result in a collection of separate performance logs that logically form a unit but cannot intuitively be investigated together with established performance tools. In this paper, we present a tool library that can be used to combine separate performance logs and separately recorded metrics into one single performance log, enabling investigation of such performance data as a unit. Use cases from Big Data processing and AI show the broad applicability of our approach.

Deep artificial neural networks (DNNs) trained through backpropagation provide effective models of the mammalian visual system, accurately capturing the hierarchy of neural responses through primary visual cortex to inferior temporal cortex (IT). However, the ability of these networks to explain representations in higher cortical areas is relatively lacking and considerably less well researched. For example, DNNs have been less successful as a model of the egocentric to allocentric transformation embodied by circuits in retrosplenial and posterior parietal cortex. We describe a novel scene perception benchmark inspired by a hippocampal dependent task, designed to probe the ability of DNNs to transform scenes viewed from different egocentric perspectives. Using a network architecture inspired by the connectivity between temporal lobe structures and the hippocampus, we demonstrate that DNNs trained using a triplet loss can learn this task. Moreover, by enforcing a factorized latent space, we can split information propagation into ``what'' and ``where'' pathways, which we use to reconstruct the input. This allows us to beat the state-of-the-art for unsupervised object segmentation on the CATER and MOVi-A,B,C benchmarks.

The Webis MS MARCO Anchor Text 2022 dataset enriches Version 1 and 2 of the document collection of MS MARCO with anchor text extracted from six Common Crawl snapshots. The six Common Crawl snapshots cover the years 2016 to 2021 (between 1.7-3.4 billion documents each). We sampled 1,000 anchor texts for documents with more than 1,000 anchor texts at random and all anchor texts for documents with less than 1,000 anchor texts (this sampling yields that all anchor text is included for 94% of the documents in Version 1 and 97% of documents for Version 2). Overall, the MS MARCO Anchor Text 2022 dataset enriches 1,703,834 documents for Version 1 and 4,821,244 documents for Version 2 with anchor text. Cleaned versions of the MS MARCO Anchor Text 2022 dataset are available in ir_datasets, Zenodo and Hugging Face. The raw dataset with additional information and all metadata for the extracted anchor texts (roughly 100GB) is available on Hugging Face and files.webis.de. The details of the construction of the Webis MS MARCO Anchor Text 2022 dataset are described in the associated paper. If you use this dataset, please cite @InProceedingsfroebe:2022a, address = Berlin Heidelberg New York, author = Maik Fr``obe and Sebastian G''unther and Maximilian Probst and Martin Potthast and Matthias Hagen, booktitle = Advances in Information Retrieval. 44th European Conference on IR Research (ECIR 2022), editor = Matthias Hagen and Suzan Verberne and Craig Macdonald and Christin Seifert and Krisztian Balog and Kjetil Norvrag and Vinay Setty, month = apr, publisher = Springer, series = Lecture Notes in Computer Science, site = Stavanger, Norway, title = The Power of Anchor Text in the Neural Retrieval Era, year = 2022

Barcode-based tracking of individuals is revolutionizing animal behavior studies, but further progress hinges on whether in addition to determining an individual's location, specific behaviors can be identified and monitored. We achieve this goal using information from the barcodes to identify tightly bounded image regions that potentially show the behavior of interest. These image regions are then analyzed with convolutional neural networks to verify that the behavior occurred. When applied to a challenging test case, detecting social liquid transfer (trophallaxis) in the honey bee hive, this approach yielded a 67% higher sensitivity and an 11% lower error rate than the best detector for honey bee trophallaxis so far. We were furthermore able to automatically detect whether a bee donates or receives liquid, which previously required manual observations. By applying our trophallaxis detector to recordings from three honey bee colonies and performing simulations, we discovered that liquid exchanges among bees generate two distinct social networks with different transmission capabilities. Finally, we demonstrate that our approach generalizes to detecting other specific behaviors. We envision that its broad application will enable automatic, high-resolution behavioral studies that address a broad range of previously intractable questions in evolutionary biology, ethology, neuroscience, and molecular biology.

Recent advances in large language models have enabled the development of viable generative retrieval systems. Instead of a traditional document ranking, generative retrieval systems often directly return a grounded generated text as a response to a query. Quantifying the utility of the textual responses is essential for appropriately evaluating such generative ad hoc retrieval. Yet, the established evaluation methodology for ranking-based ad hoc retrieval is not suited for the reliable and reproducible evaluation of generated responses. To lay a foundation for developing new evaluation methods for generative retrieval systems, we survey the relevant literature from the fields of information retrieval and natural language processing, identify search tasks and system architectures in generative retrieval, develop a new user model, and study its operationalization.

We propose to use image captions from the Web as a previously underutilized resource for paraphrases (i.e., texts with the same ``message'') and to create and analyze a corresponding dataset. When an image is reused on the Web, an original caption is often assigned. We hypothesize that different captions for the same image naturally form a set of mutual paraphrases. To demonstrate the suitability of this idea, we analyze captions in the English Wikipedia, where editors frequently relabel the same image for different articles. The paper introduces the underlying mining technology, the resulting Wikipedia-IPC dataset, and compares known paraphrase corpora with respect to their syntactic and semantic paraphrase similarity to our new resource. In this context, we introduce characteristic maps along the two similarity dimensions to identify the style of paraphrases coming from different sources. An annotation study demonstrates the high reliability of the algorithmically determined characteristic maps.

It is challenging to implement Kernel methods, if the data sources are distributed and cannot be joined at a trusted third party for privacy reasons. It is even more challenging, if the use case rules out privacy-preserving approaches that introduce noise. An example for such a use case is machine learning on clinical data. To realize exact privacy preserving computation of kernel methods, we propose FLAKE, a Federated Learning Approach for KErnel methods on horizontally distributed data. With FLAKE, the data sources mask their data so that a centralized instance can compute a Gram matrix without compromising privacy. The Gram matrix allows to calculate many kernel matrices, which can be used to train kernel-based machine learning algorithms such as Support Vector Machines. We prove that FLAKE prevents an adversary from learning the input data or the number of input features under a semi-honest threat model. Experiments on clinical and synthetic data confirm that FLAKE is outperforming the accuracy and efficiency of comparable methods. The time needed to mask the data and to compute the Gram matrix is several orders of magnitude less than the time a Support Vector Machine needs to be trained. Thus, FLAKE can be applied to many use cases.

With the growth of traffic on inland waterways, autonomous driving technologies for vessels will gain increasing significance to ensure traffic flow and safety. Inspired by car-following models for road traffic, which demonstrated their strength to reduce stop-and-go waves and increase efficiency and safety, we propose a vessel-following model for inland waterways based on deep reinforcement learning (RL). Our model is trained under consideration of realistic vessel dynamics and environmental influences, such as varying stream velocity and river profile, and with a reward function favoring observed following behavior and comfort. Aiming at high generalization capabilities, we propose a training environment that uses stochastic processes to model leading the trajectory and river dynamics. Our model demonstrated safe and comfortable driving in different unseen scenarios, including realistic vessel-following on the Middle Rhine. In comparison with an existing model, our model was able to early anticipate safety–critical situations, resulting in higher safety while maintaining comparable efficiency and comfort. In further experiments, the proposed approach demonstrated its potential to dampen traffic oscillations and reduce congestion by using a sequence of followers.

We analyze the properties of the Maximum Likelihood (ML) estimator when the underlying log-likelihood function is numerically maximized with the so-called zig-zag algorithm. By splitting the parameter vector into sub-vectors, the algorithm maximizes the log-likelihood function alternatingly with respect to one sub-vector while keeping the others constant. For situations when the algorithm is initialized with a consistent estimator and is iterated sufficiently often, we establish the asymptotic equivalence of the zig-zag estimator and the “infeasible” ML estimator being numerically approximated. This result gives guidance for practical implementations. We illustrate how to employ the algorithm in different estimation problems, such as in a vine copula model and a vector autoregressive moving average model. The accuracy of the estimator is illustrated through simulations. Finally, we demonstrate the usefulness of our results in an application, where the Bitcoin heating 2017 is analyzed by a dynamic conditional correlation model.

AIMS: Artificial intelligence (AI) provides a powerful tool to extract information from digitised histopathology whole slide images. During the last 5 years, academic and commercial actors have developed new technical solutions for a diverse set of tasks, including tissue segmentation, cell detection, mutation prediction, prognostication and prediction of treatment response. In the light of limited overall resources, it is presently unclear for researchers, practitioners and policymakers which of these topics are stable enough for clinical use in the near future and which topics are still experimental, but worth investing time and effort into. METHODS AND RESULTS: To identify potentially promising applications of AI in pathology, we performed an anonymous online survey of 75 computational pathology domain experts from academia and industry. Participants enrolled in 2021 were queried about their subjective opinion on promising and appealing subfields of computational pathology with a focus upon solid tumours. The results of this survey indicate that the prediction of treatment response directly from routine pathology slides is regarded as the most promising future application. This item was ranked highest in the overall analysis and in subgroups by age and professional background. Furthermore, prediction of genetic alterations, gene expression and survival directly from routine pathology images scored consistently high throughout subgroups. CONCLUSIONS: Together, these data demonstrate a possible direction for the development of computational pathology systems in clinical, academic and industrial research in the near future.

Electroactive polymers are frequently used in engineering applications due to their ability to change their shape and properties under the influence of an electric field. This process also works vice versa, such that mechanical deformation of the material induces an electric field in the EAP device. This specific behavior makes such materials highly attractive for the construction of actuators and sensors in various application areas. The electromechanical behaviour of electroactive polymers can be described by a third-order coupling tensor, which represents the sensitivity of mechanical stresses concerning the electric field, i.e., it establishes a relation between a second-order and a first-order tensor field. Due to this coupling tensor's complexity and the lack of meaningful visualization methods for third-order tensors in general, an interpretation of the tensor is rather difficult. Thus, the central engineering research question that this contribution deals with is a deeper understanding of electromechanical coupling by analyzing the third-order coupling tensor with the help of specific visualization methods. Starting with a deviatoric decomposition of the tensor, the multipoles of each deviator are visualized, which allows a first insight into this highly complex third-order tensor. In the present contribution, four examples, including electromechanical coupling, are simulated within a finite element framework and subsequently analyzed using the tensor visualization method.

Assessment is a fundamental factor in monitoring the learning process of students and therefore an essential component of effective teaching and learning in the online environment. In the course of the (corona-induced) wave of digitalization, the new and different forms of assessment present us with new challenges. The book focuses on these new forms of digital assessments and highlights effective practices and opportunities associated with conducting assessments in digital and hybrid learning environments.

Dieser Kurs ist ein fakultätsübergreifendes Qualifizierungsprogramm für Lehrende, das im Rahmen des Projektes „Learning Analytics – Studierende im Fokus“ entwickelt wurde. Der MOOC setzt sich mit dem Einsatz von Learning Analytics in der Hochschullehre auseinander und soll eine studierendenorientierte Lehrentwicklung unterstützen. Möglichkeiten und Grenzen der Technologie in hochschuldidaktischem Kontext oder der Einsatz von Methoden und Tools in Learning Analytics werden differenziert betrachtet. Dazu gehören etwa die Analyse von Lernaktivitäten, die Identifikation von Lernbedürfnissen und -problemen sowie die Visualisierung von Lernaktivitäten. Die Teilnehmer*innen erfahren, wie Daten aus verschiedenen Quellen gesammelt und aufbereitet werden und wie sie Learning Analytics in ihre Lehrmethoden und -strategien einbinden können, um die Qualität der Lehre zu verbessern. Das soll ihnen erlauben, die Daten umfassend zu interpretieren und didaktisch relevante Schlussfolgerungen zu ziehen. Zentral im Einsatz von Learning Analytics in der Hochschullehre sind zudem Datenschutz- und ethische Aspekte, die in diesem Kurs ebenfalls behandelt werden.Dieser MOOC richtet sich einerseits an Hochschullehrende, die durch den Einsatz von Learning Analytics die Lernprozesse der Studierenden besser verstehen und auf dieser Grundlage ihre Lehre optimieren möchten. Andererseits werden Hochschuldidaktikerinnen und Hochschuldidaktiker sowie Weiterbildnerinnen und Weiterbildner angesprochen, welche Lehrende beim Einsatz von Learning Analytics in der Lehre unterstützen.

We present a C++14 library for performance portability of scientific computing codes across CPU and GPU architectures. Our library combines generic data structures like vectors, multi-dimensional arrays, maps, graphs, and sparse grids with basic, reusable algorithms like convolutions, sorting, prefix sum, reductions, and scan. The memory layout of the data structures is adapted at compile-time using tuples with optional memory mirroring between CPU and GPU. We combine this transparent memory mapping with generic algorithms under two alternative programming interfaces: a CUDA-like kernel interface for multi-core CPUs, Nvidia GPUs, and AMD GPUs, as well as a lambda interface. We validate and benchmark the presented library using micro-benchmarks, showing that the abstractions introduce negligible performance overhead, and we compare performance against the current state of the art.

The computation of reliable, chemically correct atom maps from educt/product pairs has turned out to be a difficult problem in cheminformatics because the chemically correct solution is not necessarily an optimal solution for combinatorial formulations such as maximum common subgraph problems. As a consequence, competing models have been devised and compared in extensive benchmarking studies. Due to isomorphisms among products and educts it is not immediately obvious, however, when two atom maps for a given educt/product pairs are the same. We formalize here the equivalence of atom maps and show that equivalence of atom maps is in turn equivalent to the isomorphism of labeled auxiliary graphs. In particular, we demonstrate that Fujita's Imaginary Transition State can be used for this purpose. Numerical experiments show that practical feasibility. Generalizations to the equivalence of subgraph matches, double pushout graph transformation rules, and mechanisms of multi-step reactions are discussed briefly.

The ImageCLEF evaluation campaign was integrated with CLEF (Conference and Labs of the Evaluation Forum) for more than 20 years and represents a Multimedia Retrieval challenge aimed at evaluating the technologies for annotation, indexing, and retrieval of multimodal data. Thus, it provides information access to large data collections in usage scenarios and domains such as medicine, argumentation and content recommendation. ImageCLEF 2024 has four main tasks: (i) a Medical task targeting automatic image captioning for radiology images, synthetic medical images created with Generative Adversarial Networks (GANs), Visual Question Answering and medical image generation based on text input, and multimodal dermatology response generation; (ii) a joint ImageCLEF-Touché task Image Retrieval/Generation for Arguments to convey the premise of an argument, (iii) a Recommending task addressing cultural heritage content-recommendation, and (iv) a joint ImageCLEF-ToPicto task aiming to provide a translation in pictograms from natural language. In 2023, participation increased by 67% with respect to 2022 which reveals its impact on the community.

Purpose: The interpretation of image data plays a critical role during acute brain stroke diagnosis, and promptly defining the requirement of a surgical intervention will drastically impact the patient's outcome. However, determining stroke lesions purely from images can be a daunting task. Many studies proposed automatic segmentation methods for brain stroke lesions from medical images in different modalities, though heretofore results do not satisfy the requirements to be clinically reliable. We investigate the segmentation of brain stroke lesions using a geometric deep learning model that takes advantage of the intrinsic interconnected diffusion features in a set of multi-modal inputs consisting of computer tomography (CT) perfusion parameters. Approach: We propose a geometric deep learning model for the segmentation of ischemic stroke brain lesions that employs spline convolutions and unpooling/pooling operators on graphs to excerpt graph-structured features in a fully convolutional network architecture. In addition, we seek to understand the underlying principles governing the different components of our model. Accordingly, we structure the experiments in two parts: an evaluation of different architecture hyperparameters and a comparison with state-of-the-art methods. Results: The ablation study shows that deeper layers obtain a higher Dice coefficient score (DCS) of up to 0.3654. Comparing different pooling and unpooling methods shows that the best performing unpooling method is the proportional approach, yet it often smooths the segmentation border. Unpooling achieves segmentation results more adapted to the lesion boundary corroborated with systematic lower values of Hausdorff distance. The model performs at the level of state-of-the-art models without optimized training methods, such as augmentation or patches, with a DCS of 0.4553$pm$0.0031. Conclusions: We proposed and evaluated an end-to-end trainable fully convolutional graph network architecture using spline convolutional layers for the ischemic stroke lesion prediction. We propose a model that employs graph-based operations to predict acute stroke brain lesions from CT perfusion parameters. Our results prove the feasibility of using geometric deep learning to solve segmentation problems, and our model shows a better performance than other models evaluated. The proposed model achieves improved metric values for the DCS metric, ranging from 8.61% to 69.05%, compared with other models trained under the same conditions. Next, we compare different pooling and unpooling operations in relation to their segmentation results, and we show that the model can produce segmentation outputs that adapt to irregular segmentation boundaries when using simple heuristic unpooling operations.

The bandwidth improvement provided by high-bandwidth memory (HBM), and the capability of FPGAs to customize the processing and memory hierarchy, results in a considerable performance increase for memory-intensive work-loads such as graph processing, sorting, machine learning, and database analytics. Modern systems integrating 3D-stacked DRAM memory can be leveraged to realize the Near-Memory Computing (NMC) paradigm by offloading some computations to accelerators placed near the HBM. Although numerous studies have investigated efficient accelerators for FPGA-HBM platforms, researchers have not proposed a systematic way for identifying which application kernels are suitable for execution near the HBM. In this article, we propose compiler support for recognizing offloading candidates without any burden on programmers. Auto-DOK analyzes an application code based on criteria derived from the hardware design goals of FPGA-HBM platforms, and automatically identifies kernels suitable for offloading. We evaluate Auto-DOK on benchmarks ranging from microbenchmarks to real-world kernels. Our results show that Auto-DOK can correctly identify kernels and input sizes suitable for execution near the HBM, and prevents slowdown caused by incorrect offloading decisions for other workloads. Moreover, Auto-DOK operates at compile time with negligible overhead and without the need for expensive profiling.

The near-data processing (NDP) paradigm has recently gained popularity as a promising method for mitigating the memory wall challenges of future computing systems. Modern 3D-stacked DRAM memories can be leveraged to prevent unnecessary data movement between the main memory and the CPU. However, identifying which type of NDP system, in terms of processing elements and the type of high performance 3D memory used, is suitable for a given application is not a trivial task. Understanding the interactions between modern workloads and the memory subsystem requires a lot of effort. Each memory type presents its unique advantages and shortcomings. In addition, memory access patterns vary greatly across applications. Therefore, the performance of a given application on a given memory type is difficult to intuitively predict. There is no specific NDP system or memory type that can provide high performance for every application. We propose NDP-RANK, a machine learning framework that can efficiently decide which NDP system is suitable for an application. NDP-RANK’s performance prediction is based on an input set of application and microarchitecture characteristics. We build machine learning models that can be used to conduct design-space exploration of NDP systems and accurately predict performance of previously unseen applications on several NDP architectures. Our models are orders of magnitude faster than architectural simulation. They can accurately predict performance with coefficients of determination ranging between 0.89 and 0.96, and root mean square errors ranging between 0.04 and 0.16.

The speed of modern computing systems has improved significantly, thanks to advances in CMOS technology. However, the memory bandwidth of DRAM has not kept pace with these improvements in terms of latency and energy consumption, which is known as the memory wall [1]. FPGAs with high-bandwidth memory (HBM) provide significantly improved performance on memory-intensive tasks, such as graph processing and machine learning. By leveraging 3D-stacked DRAM memory on FPGAs, it is possible to realize the Near-Memory Computing (NMC) paradigm, which involves offloading some kernels to be processed close to the memory. While there have been many studies on NMC accelerators, there is no established method for determining which application kernels are suitable for execution near the HBM. To fully realize the potential of FPGA-HBM architectures, it is important to identify offloading candidates without relying on programmers' knowledge. However, this is a non-trivial task due to the complexity of modern applications. To address this issue, we propose a compiler-assisted tool-flow for the automatic selection of kernels to be offloaded.

The concept of near-memory computing (NMC) has emerged as a promising solution to address the memory wall challenges faced by future computing architectures. By utilizing modern systems that integrate 3D-stacked DRAM memory, the NMC paradigm minimizes unnecessary data movement between the memory subsystem and the CPU. FPGA vendors have incorporated 3D-stacked memories into their products to meet the increasing bandwidth requirements of memory-intensive applications, enabling FPGAs to compete with GPU solutions in terms of speed and energy efficiency. Recent NMC proposals focus on different data processing workloads, including graph processing and machine learning. This work addresses the research questions of how to leverage the full bandwidth of 3D-stacked high-bandwidth memory and how to facilitate the adoption of the near-memory computing paradigm.

This system demonstration presents Nemo, a new logic programming engine with a focus on reliability and performance. Nemo is built for data-centric analytic computations, modelled in a fully declarative Datalog dialect. Its scalability for these tasks matches or exceeds that of leading Datalog systems. We demonstrate uses in reasoning with knowledge graphs and ontologies with 10^5 to 10^8 input facts, all on a laptop. Nemo is written in Rust and available as a free and open source tool.

The field of computer vision is steadily growing in its complexity and application areas. FPGAs have shown that they can meet the growing demands for performance and energy efficiency. However, their programmability is a major challenge for software programmers. With OpenVX a standard for cross platform acceleration of computer vision applications exists. Existing OpenVX FPGA frameworks often contain non-standard constructs and consider either fixed processor architectures or specialized non-adaptive accelerators. Therefore, we propose ArcvaVX, a framework that generates a runtime-adaptive vision architecture from OpenVX applications, which is performant and flexible. It (1) verifies the user implemented OpenVX applications, partitions them into task graphs and creates their meta-data (2) maps these tasks to physical nodes, creates a schedule, and clusters and places the nodes within a partition-based topology (3) creates the hardware architecture, including additional components required to generate a runtime-adaptive system. These components contain runtime configurable network adapters that can prevent deadlocks, a controller for direct memory access, and a manager that configures both and maintains the schedule. The architecture is designed for applications with high data rates and low synchronization overhead. The evaluation shows a low latency overhead of 0.006% added by the architecture, while resource consumption is more than halved compared to a design consisting only of accelerators.

Metagenomics became a standard strategy to comprehend the functional potential of microbial communities, including the human microbiome. Currently, the number of metagenomes in public repositories is increasing exponentially. The Sequence Read Archive (SRA) and the MG-RAST are the two main repositories for metagenomic data. These databases allow scientists to reanalyze samples and explore new hypotheses. However, mining samples from them can be a limiting factor, since the metadata available in these repositories is often misannotated, misleading, and decentralized, creating an overly complex environment for sample reanalysis. The main goal of the HumanMetagenomeDB is to simplify the identification and use of public human metagenomes of interest. HumanMetagenomeDB version 1.0 contains metadata of 69 822 metagenomes. We standardized 203 attributes, based on standardized ontologies, describing host characteristics (e.g. sex, age and body mass index), diagnosis information (e.g. cancer, Crohn's disease and Parkinson), location (e.g. country, longitude and latitude), sampling site (e.g. gut, lung and skin) and sequencing attributes (e.g. sequencing platform, average length and sequence quality). Further, HumanMetagenomeDB version 1.0 metagenomes encompass 58 countries, 9 main sample sites (i.e. body parts), 58 diagnoses and multiple ages, ranging from just born to 91 years old. The HumanMetagenomeDB is publicly available at https://webapp.ufz.de/hmgdb/.

AbstractThe governments' mitigation measures to fight the COVID-19 pandemic are unprecedented in our post-war history. For overcoming this crisis, citizens were expected to act in compliance with these measures in order to control the spread of the virus and keep public health systems functional. This call for protecting the public health at the same time confronted citizens with several and severe limitations of their democratic freedoms and rights: confinement, restriction on freedoms of movement, religion, specific provisions for public protest and finally also limitations to the right of education by school closures. This paper analyzes how citizens perceive the threat the COVID-19 pandemic and especially the mitigation measures posed for democracy. We assume that pandemic waves and pandemic fatigue have an impact on the perception of threat. To see the overall societal picture, we exploit a large-scale archive of online discourse on Twitter out of which we extract democracy-related discourse with the same temporal and geospatial coverage for our investigation. From that data source, we apply computational methods to extract time series data reflecting aggregated opinions and their evolution over time concerned with the correlation of attitudes towards democracy. We them move deeper using a longitudinal panel survey we conducted in November/December 2020, March/April 2021, and July/August 2021. to have a view of the relationship between citizens' socio-economic status and basic political attitudes. Our multi-method analysis bases on the German case and covers the period from December 2020 to August 2021.

textcopyright 2021 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).Idiosyncrasies in human writing styles make it difficult to develop systems for authorship identification that scale well across individuals. In this year's edition of PAN, the authorship identification track focused on open-set authorship verification, so that systems are applied to unknown documents by previously unseen authors in a new domain. As in the previous year, the sizable materials for this campaign were sampled from English-language fanfiction. The calibration materials handed out to the participants were the same as last year, but a new test set was compiled with authors and fandom domains not present in any of the previous datasets. The general setup of the task did not change, i.e., systems still had to estimate the probability of a pair of documents being authored by the same person. We attracted 13 submissions by 10 international teams, which were compared to three complementary baselines, using five diverse evaluation metrics. Post-hoc analyses show that systems benefitted from the abundant calibration materials and were well-equipped to handle the open-set scenario: Both the top-performing approach and the highly competitive cohort of runner-ups presented surprisingly strong verifiers. We conclude that, at least within this specific text variety, (large-scale) open-set authorship verification is not necessarily or inherently more difficult than a closed-set setup, which offers encouraging perspectives for the future of the field.

Quasi-best match graphs (qBMGs) are a hereditary class of directed, properly vertex-colored graphs. They arise naturally in mathematical phylogenetics as a generalization of best match graphs, which formalize the notion of evolutionary closest relatedness of genes (vertices) in multiple species (vertex colors). They are explained by rooted trees whose leaves correspond to vertices. In contrast to BMGs, qBMGs represent only best matches at a restricted phylogenetic distance. We provide characterizations of qBMGs that give rise to polynomial-time recognition algorithms and identify the BMGs as the qBMGs that are color-sink-free. Furthermore, two-colored qBMGs are characterized as directed graphs satisfying three simple local conditions, two of which have appeared previously, namely bi-transitivity in the sense of Das et al. (2021) and a hierarchy-like structure of out-neighborhoods, i.e., N(x)∩N(y)∈N(x),N(y),0̸ for any two vertices x and y. Further results characterize qBMGs that can be explained by binary phylogenetic trees.

AI-based IoT applications relying on heavy-load deep learning algorithms like CNNs challenge IoT devices that are restricted in energy or processing capabilities. Edge computing offers an alternative by allowing the data to get offloaded to so-called edge servers with hardware more powerful than IoT devices and physically closer than the cloud. However, the increasing complexity of data and algorithms and diverse conditions make even powerful devices, such as those equipped with FPGAs, insufficient to cope with the current demands. In this case, optimizations in the algorithms, like pruning and early-exit, are mandatory to reduce the CNNs computational burden and speed up inference processing. With that in mind, we propose ExpOL, which combines the pruning and early-exit CNN optimizations in a system-level FPGA-based IoT-Edge design space exploration. Based on a user-defined multi-target optimization, ExpOL delivers designs tailored to specific application environments and user needs. When evaluated against state-of-the-art FPGA-based accelerators (either local or offloaded), designs produced by ExpOL are more power-efficient (by up to 2times) and process inferences at higher user quality of experience (by up to 12.5%).

The challenge of processing heavy-load ML tasks, particularly CNN-based ones at resource-constrained IoT devices, has encouraged the use of edge servers. The edge offers performance levels higher than the end devices and better latency and security levels than the Cloud. On top of that, the rising complexity of ML applications, the ever-increasing number of connected devices, and the current demands for energy efficiency require optimizing such CNN models. Pruning and early-exit are notable optimizations that have been successfully used to alleviate the computational cost of inference. However, these optimizations have not yet been exploited simultaneously: while pruning is usually applied at design time, which involves retraining the CNN before deployment, early-exit is inherently dynamic. In this work, we propose AdaPEx, a framework that exploits the intrinsic reconfigurable FPGA capabilities so both can be cooperatively employed. AdaPEx first explores the trade-off between pruning and early-exit at design-time, creating a design space never exploited in the state-of-the-art. Then, AdaPEx applies FPGA reconfiguration as a means to enable the combined use of pruning and early-exit dynamically. At runtime, this allows matching the inference processing to the current edge conditions and a user-configurable accuracy threshold. In a smart IoT application, AdaPEx processes up to 1.32× more inferences and improves EDP by up to 2.55× over the state-of-the-art FPGA-based FINN accelerator.

We develop an interactive approach for analyzing multi-field tensor data from simulations in close collaboration with domain scientists. Our approach is based on extensive application analysis and built around a multi-field clustering addressing multiple user-defined quantities which were required by the domain scientists. Established techniques like linked views complement the approach to support reasoning while offering an overview and detailed insight into the multi-field tensor data. Further, we include an evaluation containing a real-world use case and a user study with domain scientists to demonstrate the usefulness compared to existing tools.

Manual point cloud registration is often a crucial step during the mapping of 3D point clouds and usually performed on a conventional desktop setup with mouse interaction. Since 3D point clouds are inherently spatial, these 2D applications suffer from impaired depth perception and inconvenient interaction. Nonetheless, there are few efforts to improve the usability of these applications. To address this, we propose an alternative setup, consisting of a stereoscopic display and an external hand tracker, allowing for enhanced depth perception and natural interaction without the need for body-worn devices or handheld controllers. We developed interaction techniques for point cloud alignment in 3D space, including visual feedback during alignment, and implemented a proof-of-concept prototype in the context of a surgical use case. We describe the use case, design and implementation of our concepts and outline future work. Herewith we provide a user-centered alternative to desktop applications for manual point cloud registration.

A special form of failure in impact loaded Fibre-Reinforced Composites (FRP) structures is delamination, in which individual layers of a laminate get separated from one another. In contrast to the continuum mechanically formulated models of damage mechanics, the description of delamination processes is based on concepts of fracture mechanics. Here, delamination initiation is due to interlaminar stresses, whereby the tolerable interlaminar shear stresses can be increased by a simultaneous through thickness compression (TTC). Furthermore, an increase in the critical energy release rate with increasing out-of-plane compressive load is described. Failure to consider the compressive superposition can lead to an overestimation of the delamination failure in impact loaded FRP structures such as three-point bending beams.

Real-world deployment of reliable object detectors is crucial for applications such as autonomous driving. However, general-purpose object detectors like Faster R-CNN are prone to providing overconfident predictions for outlier objects. Recent outlier-aware object detection approaches estimate the density of instance-wide features with class-conditional Gaussians and train on synthesized outlier features from their low-likelihood regions. However, this strategy does not guarantee that the synthesized outlier features will have a low likelihood according to the other class-conditional Gaussians. We propose a novel outlier-aware object detection framework that distinguishes outliers from inlier objects by learning the joint data distribution of all inlier classes with an invertible normalizing flow. The appropriate sampling of the flow model ensures that the synthesized outliers have a lower likelihood than inliers of all object classes, thereby modeling a better decision boundary between inlier and outlier objects. Our approach significantly outperforms the state-of-the-art for outlier-aware object detection on both image and video datasets. Code available at https://github.com/nish03/FFS

INTRODUCTION: Dementia syndromes can be difficult to diagnose. We aimed at building a classifier for multiple dementia syndromes using magnetic resonance imaging (MRI). METHODS: Atlas-based volumetry was performed on T1-weighted MRI data of 426 patients and 51 controls from the multi-centric German Research Consortium of Frontotemporal Lobar Degeneration including patients with behavioral variant frontotemporal dementia, Alzheimer's disease, the three subtypes of primary progressive aphasia, i.e., semantic, logopenic and nonfluent-agrammatic variant, and the atypical parkinsonian syndromes progressive supranuclear palsy and corticobasal syndrome. Support vector machine classification was used to classify each patient group against controls (binary classification) and all seven diagnostic groups against each other in a multi-syndrome classifier (multiclass classification). RESULTS: The binary classification models reached high prediction accuracies between 71 and 95% with a chance level of 50%. Feature importance reflected disease-specific atrophy patterns. The multi-syndrome model reached accuracies of more than three times higher than chance level but was far from 100%. Multi-syndrome model performance was not homogenous across dementia syndromes, with better performance in syndromes characterized by regionally specific atrophy patterns. Whereas diseases generally could be classified vs controls more correctly with increasing severity and duration, differentiation between diseases was optimal in disease-specific windows of severity and duration. DISCUSSION: Results suggest that automated methods applied to MR imaging data can support physicians in diagnosis of dementia syndromes. It is particularly relevant for orphan diseases beside frequent syndromes such as Alzheimer's disease.

We present the first privacy-preserving approach for stress detection from wrist-worn wearables based on the Time-Series Classification Transformer (TSCT) architecture and incorporating Differential Privacy (DP) to ensure provable privacy guarantees. The non-private baseline results prove the TSCT to be an effective model for the given task. Our DP experiments then show that the private models suffer from reduced utility but can still be used for reliable stress detection depending on the application. Our proposed approach has potential applications in smart health, where it can be used to monitor smartwatch users' stress levels without compromising their privacy and provide timely interventions or suggestions to prevent adverse health outcomes. Another primary contribution is our evaluation, which studies and shows negative effects of DP regarding model training. The results of this work provide perspectives for future research and applications whenever the fields of stress detection and data privacy intervene.

A single experimental method alone often fails to provide the resolution, accuracy, and coverage needed to model integral membrane proteins (IMPs). Integrating computation with experimental data is a powerful approach to supplement missing structural information with atomic detail. We combine RosettaNMR with experimentally-derived paramagnetic NMR restraints to guide membrane protein structure prediction. We demonstrate this approach using the disulfide bond formation protein B (DsbB), an $alpha$-helical IMP. Here, we attached a cyclen-based paramagnetic lanthanide tag to an engineered non-canonical amino acid (ncAA) using a copper-catalyzed azide-alkyne cycloaddition (CuAAC) click chemistry reaction. Using this tagging strategy, we collected 203 backbone HN pseudocontact shifts (PCSs) for three different labeling sites and used these as input to guide de novo membrane protein structure prediction protocols in Rosetta. We find that this sparse PCS dataset combined with 44 long-range NOEs as restraints in our calculations improves structure prediction of DsbB by enhancements in model accuracy, sampling, and scoring. The inclusion of this PCS dataset improved the C$alpha$-RMSD transmembrane segment values of the best-scoring and best-RMSD models from 9.57 AA and 3.06 AA (no NMR data) to 5.73 AA and 2.18 AA, respectively.

A key ingredient of computational argumentation in AI is the generation of arguments in favor of or against claims under scrutiny. In this paper we look at the complexity of argument construction and reasoning in the prominent structured formalism of assumption-based argumentation (ABA). We point out that reasoning in ABA by means of constructing an abstract argumentation framework (AF) gives rise to two main sources of complexity: (i) constructing the AF and (ii) reasoning within the constructed graph. Since both steps are intractable in general, it is no surprise that the best performing state-of-the-art ABA reasoners skip the instantiation procedure entirely and perform tasks directly on the input knowledge base. Driven by this observation, we identify and study atomic and symmetric ABA, two ABA fragments that preserve the expressive power of general ABA, and that can be utilized to have milder complexity in the first or second step. We show that using atomic ABA allows for an instantiation procedure for general ABA leading to polynomially-bounded AFs and that symmetric ABA can be used to create AFs that have mild complexity to reason on. By an experimental evaluation, we show that using the former approach with modern AF solvers can be competitive with state-of-the-art ABA solvers, improving on previous AF instantiation approaches that are hindered by intractable argument construction.

In the autonomous driving field, fusion of human knowledge into Deep Reinforcement Learning (DRL) is often based on the human demonstration recorded in a simulated environment. This limits the generalization and the feasibility of application in real-world traffic. We propose a two-stage DRL method to train a car-following agent, that modifies the policy by leveraging the real-world human driving experience and achieves performance superior to the pure DRL agent. Training a DRL agent is done within CARLA framework with Robot Operating System (ROS). For evaluation, we designed different driving scenarios to compare the proposed two-stage DRL car-following agent with other agents. After extracting the “good” behavior from the human driver, the agent becomes more efficient and reasonable, which makes this autonomous agent more suitable to Human–Robot Interaction (HRI) traffic.

To achieve fully autonomous driving, vehicles must be capable of continuously performing various driving tasks, including lane keeping and car following, both of which are fundamental and well-studied driving ones. However, previous studies have mainly focused on individual tasks, and car following tasks have typically relied on complete leader-follower information to attain optimal performance. To address this limitation, we propose a vision-based deep reinforcement learning (DRL) agent that can simultaneously perform lane keeping and car following maneuvers. To evaluate the performance of our DRL agent, we compare it with a baseline controller and use various performance metrics for quantitative analysis. Furthermore, we conduct a real-world evaluation to demonstrate the Sim2Real transfer capability of the trained DRL agent. To the best of our knowledge, our vision-based car following and lane keeping agent with Sim2Real transfer capability is the first of its kind. We have made the codes and the videos of the simulation and real-world evaluation accessible online11Code and videos are available on: https://github.com/DailyL/Sim2Realautonomous_vehicle.

The BigCode community, an open-scientific collaboration working on the responsible development of Large Language Models for Code (Code LLMs), introduces StarCoder and StarCoderBase: 15.5B parameter models with 8K context length, infilling capabilities and fast large-batch inference enabled by multi-query attention. StarCoderBase is trained on 1 trillion tokens sourced from The Stack, a large collection of permissively licensed GitHub repositories with inspection tools and an opt-out process. We fine-tuned StarCoderBase on 35B Python tokens, resulting in the creation of StarCoder. We perform the most comprehensive evaluation of Code LLMs to date and show that StarCoderBase outperforms every open Code LLM that supports multiple programming languages and matches or outperforms the OpenAI code-cushman-001 model. Furthermore, StarCoder outperforms every model that is fine-tuned on Python, can be prompted to achieve 40% pass@1 on HumanEval, and still retains its performance on other programming languages. We take several important steps towards a safe open-access model release, including an improved PII redaction pipeline and a novel attribution tracing tool, and make the StarCoder models publicly available under a more commercially viable version of the Open Responsible AI Model license.

G protein-coupled receptors (GPCRs) are the largest class of drug targets and undergo substantial conformational changes in response to ligand binding. Despite recent progress in GPCR structure determination, static snapshots fail to reflect the conformational space of putative binding pocket geometries to which small molecule ligands can bind. In comparative modeling of GPCRs in the absence of a ligand, often a shrinking of the orthosteric binding pocket is observed. However, the exact prediction of the flexible orthosteric binding site is crucial for adequate structure-based drug discovery. In order to improve ligand docking and guide virtual screening experiments in computer-aided drug discovery, we developed RosettaGPCRPocketSize. The algorithm creates a conformational ensemble of biophysically realistic conformations of the GPCR binding pocket between the TM bundle, which is consistent with a knowledge base of expected pocket geometries. Specifically, tetrahedral volume restraints are defined based on information about critical residues in the orthosteric binding site and their experimentally observed range of C$alpha$-C$alpha$-distances. The output of RosettaGPCRPocketSize is an ensemble of binding pocket geometries that are filtered by energy to ensure biophysically probable arrangements, which can be used for docking simulations. In a benchmark set, pocket shrinkage observed in the default RosettaGPCR was reduced by up to 80% and the binding pocket volume range and geometric diversity were increased. Compared to models from four different GPCR homology model databases (RosettaGPCR, GPCR-Tasser, GPCR-SSFE, and GPCRdb), the here-created models showed more accurate volumes of the orthosteric pocket when evaluated with respect to the crystallographic reference structure. Furthermore, RosettaGPCRPocketSize was able to generate an improved realistic pocket distribution. However, while being superior to other homology models, the accuracy of generated model pockets was comparable to AlphaFold2 models. Furthermore, in a docking benchmark using small-molecule ligands with a higher molecular weight between 400 and 700 Da, a higher success rate in creating native-like binding poses was observed. In summary, RosettaGPCRPocketSize can generate GPCR models with realistic orthosteric pocket volumes, which are useful for structure-based drug discovery applications.

As machine learning-enabled Text-to-Image (TTI) systems are becoming increasingly prevalent and seeing growing adoption as commercial services, characterizing the social biases they exhibit is a necessary first step to lowering their risk of discriminatory outcomes. This evaluation, however, is made more difficult by the synthetic nature of these systems' outputs: common definitions of diversity are grounded in social categories of people living in the world, whereas the artificial depictions of fictive humans created by these systems have no inherent gender or ethnicity. To address this need, we propose a new method for exploring the social biases in TTI systems. Our approach relies on characterizing the variation in generated images triggered by enumerating gender and ethnicity markers in the prompts, and comparing it to the variation engendered by spanning different professions. This allows us to (1) identify specific bias trends, (2) provide targeted scores to directly compare models in terms of diversity and representation, and (3) jointly model interdependent social variables to support a multidimensional analysis. We leverage this method to analyze images generated by 3 popular TTI systems (Dall-E 2, Stable Diffusion v 1.4 and 2) and find that while all of their outputs show correlations with US labor demographics, they also consistently under-represent marginalized identities to different extents. We also release the datasets and low-code interactive bias exploration platforms developed for this work, as well as the necessary tools to similarly evaluate additional TTI systems.

This paper presents Pearl, a mixed-reality approach for the analysis of human movement data in situ. As the physical environment shapes human motion and behavior, the analysis of such motion can benefit from the direct inclusion of the environment in the analytical process. We present methods for exploring movement data in relation to surrounding regions of interest, such as objects, furniture, and architectural elements. We introduce concepts for selecting and filtering data through direct interaction with the environment, and a suite of visualizations for revealing aggregated and emergent spatial and temporal relations. More sophisticated analysis is supported through complex queries comprising multiple regions of interest. To illustrate the potential of Pearl, we developed an Augmented Reality-based prototype and conducted expert review sessions and scenario walkthroughs in a simulated exhibition. Our contribution lays the foundation for leveraging the physical environment in the in-situ analysis of movement data.

We present an artificial neural network architecture, termed STENCIL-NET, for equation-free forecasting of spatiotemporal dynamics from data. STENCIL-NET works by learning a discrete propagator that is able to reproduce the spatiotemporal dynamics of the training data. This data-driven propagator can then be used to forecast or extrapolate dynamics without needing to know a governing equation. STENCIL-NET does not learn a governing equation, nor an approximation to the data themselves. It instead learns a discrete propagator that reproduces the data. It therefore generalizes well to different dynamics and different grid resolutions. By analogy with classic numerical methods, we show that the discrete forecasting operators learned by STENCIL-NET are numerically stable and accurate for data represented on regular Cartesian grids. A once-trained STENCIL-NET model can be used for equation-free forecasting on larger spatial domains and for longer times than it was trained for, as an autonomous predictor of chaotic dynamics, as a coarse-graining method, and as a data-adaptive de-noising method, as we illustrate in numerical experiments. In all tests, STENCIL-NET generalizes better and is computationally more efficient, both in training and inference, than neural network architectures based on local (CNN) or global (FNO) nonlinear convolutions.

Quinolone resistance presents a growing global health threat. We employed word-based GWAS to explore genomic data, aiming to enhance our understanding of this phenomenon. Unlike traditional variant-based GWAS analyses, this approach simultaneously captures multiple genomic factors, including single and interacting resistance mutations and genes. Analyzing a dataset of 92 genomic E. coli samples from a wastewater treatment plant in Dresden, we identified 54 DNA unitigs significantly associated with quinolone resistance. Remarkably, our analysis not only validated known mutations in gyrA and parC genes and the results of our variant-based GWAS but also revealed new (mutated) genes such as mdfA, the AcrEF-TolC multidrug efflux system, ptrB, and hisI, implicated in antibiotic resistance. Furthermore, our study identified joint mutations in 14 genes including the known gyrA gene, providing insights into potential synergistic effects contributing to quinolone resistance. These findings showcase the exceptional capabilities of word-based GWAS in unraveling the intricate genomic foundations of quinolone resistance.

Lipid molecules such as cholesterol interact with the surface of integral membrane proteins (IMP) in a mode different from drug-like molecules in a protein binding pocket. These differences are due to the lipid molecule's shape, the membrane's hydrophobic environment, and the lipid's orientation in the membrane. We can use the recent increase in experimental structures in complex with cholesterol to understand protein-cholesterol interactions. We developed the RosettaCholesterol protocol consisting of (1) a prediction phase using an energy grid to sample and score native-like binding poses and (2) a specificity filter to calculate the likelihood that a cholesterol interaction site may be specific. We used a multi-pronged benchmark (self-dock, flip-dock, cross-dock, and global-dock) of protein-cholesterol complexes to validate our method. RosettaCholesterol improved sampling and scoring of native poses over the standard RosettaLigand baseline method in 91% of cases and performs better regardless of benchmark complexity. On the $beta$2AR, our method found one likely-specific site, which is described in the literature. The RosettaCholesterol protocol quantifies cholesterol binding site specificity. Our approach provides a starting point for high-throughput modeling and prediction of cholesterol binding sites for further experimental validation.

Für ein Grundverständnis der Chancen, Risiken und Grenzen von Künstlicher Intelligenz sollten insbesondere die Methoden des Maschinellen Lernens (ML) von Schüler:innen nachvollzogen werden. Mithilfe des Workshopkonzepts „Die intelligente Museumsapp“ kann der zentrale Problemlöseprozess des ML erarbeitet werden. Es wird das Konzept des Workshops vorgestellt.

Künstliche Intelligenz (KI), insbesondere Maschinelles Lernen (ML) findet vermehrt Einzug in die Lehrpläne für den Informatikunterricht, ohne dass Lehramtsstudierende mit dem Fach Informatik dafür im Studium ausreichend qualifiziert werden. In diesem Praxisbeitrag wird ein Seminarkonzept vorgestellt, in dem Lehramtsstudierende sowohl fachlich als auch fachdidaktisch für die Vermittlung des Themengebiets ML im Unterricht befähigt werden. Im Mittelpunkt des Seminars steht die Erstellung eines Erklärvideos und eines digitalen Lernmoduls zu einem gewählten ML-Thema. Während der ersten Pilotierung mit Studierenden wird die Struktur des Seminars und die Prüfungsform positiv hervorgehoben.

With the increasing relevance of Artificial Intelligence (AI) in society, AI literacy is increasingly being integrated into K-12 education. However, students already form conceptions about modern technologies (especially AI) from their interaction with their everyday environment. In this paper, we conduct a systematic scoping review on the emerging research field of secondary school students' mental models and attitudes regarding AI. Our goal is to identify research approaches and uncover research gaps. To achieve this, we reviewed literature published within the timeframe of 2012–2022 found in the ACM, IEEE, ERIC, and Science Direct databases. In total, we identified 18 relevant papers, for which we collected and evaluated research objectives, approaches, results, and areas requiring further investigation. Our findings reveal that most of the literature examined focuses with students' attitudes toward AI and moderating variables. Additionally, there is preliminary research on the types of mental models students possess about AI. Notable research gaps include the absence of development of research instruments for surveying mental models and insufficient research into the impact of learning interventions on the construction of students' mental models.

The recent growth of Artificial Intelligence (AI) based systems considerably widespread their use in many application areas and our daily lives [28]. For instance, AI models are nowadays imbued into web search engines, self-autonomous vehicles, recommendation systems, games, and healthcare [24]. Accordingly, the demand for eXplainable AI (XAI) has risen [20] , [21] to help users cope with the growing complexity and opaqueness of the emerging generation of AI models [20] , [21]. By allowing users to perceive and make sense of the behaviors and outcomes of such models, XAI enables users of diverse levels of expertise to trust, manage, design, inspect, and develop AI models [13].

Deep representation learning is a ubiquitous part of modern computer vision. While Euclidean space has been the de facto standard manifold for learning visual representations, hyperbolic space has recently gained rapid traction for learning in computer vision. Specifically, hyperbolic learning has shown a strong potential to embed hierarchical structures, learn from limited samples, quantify uncertainty, add robustness, limit error severity, and more. In this paper, we provide a categorization and in-depth overview of current literature on hyperbolic learning for computer vision. We research both supervised and unsupervised literature and identify three main research themes in each direction. We outline how hyperbolic learning is performed in all themes and discuss the main research problems that benefit from current advances in hyperbolic learning for computer vision. Moreover, we provide a high-level intuition behind hyperbolic geometry and outline open research questions to further advance research in this direction.

Each year the International Semantic Web Conference accepts a set of Semantic Web Challenges to establish competitions that will advance the state of the art solutions in any given problem domain. The SeMantic AnsweR Type prediction task (SMART) was part of ISWC 2020 challenges. Question type and answer type prediction can play a key role in knowledge base question answering systems providing insights that are helpful to generate correct queries or rank the answer candidates. More concretely, given a question in natural language, the task of SMART challenge is, to predict the answer type using a target ontology (e.g., DBpedia or Wikidata).

Spectral Indices derived from multispectral remote sensing products are extensively used to monitor Earth system dynamics (eg vegetation dynamics, water bodies, fire regimes). The rapid increase of proposed spectral indices led to a high demand for catalogues of spectral indices and tools for their computation. However, most of these resources are either closed-source, outdated, unconnected to a catalogue or lacking a common Application Programming Interface (API). Here we present “Awesome Spectral Indices”(ASI), a standardized catalogue of spectral indices for Earth system research. ASI provides a comprehensive machine readable catalogue of spectral indices, which is linked to a Python library. ASI delivers a broad set of attributes for each spectral index, including names, formulas, and source references. The catalogue can be extended by the user community, ensuring that ASI remains current and …

Dieser Beitrag demonstriert den Einsatz von FEEDI (Feedback im Diagramm-Assessment) für Enti-ty-Relationship-Diagramme. FEEDI ist eine E-Assessment-Lösung, welche studentische ER-Diagramme automatisiert bewertet und Feedback auf eingereichte Lösungen gibt. Die Demonstration illustriert, wie Lehrende ihre Musterlösung hochladen und mittels einer GUI annotieren können. Für die Bewertung einer studentischen Lösung wandelt FEEDI die Diagramme zunächst in eine interne Graph-Repräsentation um und ermittelt mittels Graph-Matching die größtmögliche Passung. Aus dem Ergebnis des Matchings wird dann eine Bewertung sowie für fehlende bzw. falsche Aspekte der studentischen Lösung ein Feedback ausgegeben. FEEDI eignet sich als Übungs-System für Diagramme und grafische Modellierungen für Studierende im MINT-Bereich.

This repository provides supplementary material to the ICSE 2023 paper ``Analyzing the Impact of Workloads on Modeling the Performance of Configurable Software Systems''. We provide the following material: - The experimental setup, including the performance and measurement scripts. - (Aggregated) measurement data and configurations used in our analysis as well as the raw code coverage measurements. - An interactive dashboard to reproduce and reenact our analyses/findings, re-create all visualizations used in the original paper and those omitted due to space limitations. The repository is structured as follows: - accepted_paper.pdf: Camera-ready version of the original paper for reference. - coverages_raw.tar.gz: Raw coverage reports as compressed CSV files (uncrompressed: ~60 GB) - artifacts_excluding_raw_coverage.zip: aggregated measurement data, interactive dashboard, and experimental setup - README.md: A detailed documentation of all the material provided.

Accurate estimation of fuel consumption and emissions is crucial for assessing the impact of materials and stringent emission control techniques on climate change, particularly in the transportation industry, which accounts for a significant portion of global greenhouse gases and hazardous pollutants emissions. To address these concerns, the government of Canada has collected a large sensor-based dataset containing detailed information on 7384 light-duty vehicles from 2017 to 2021, with the goal of reducing CO2 emissions by 40--45% by 2030. To this end, various researchers worldwide have developed vehicle emissions and consumption models to comply with these targets and achieve the Canadian government's ambitious objectives. In this work, we propose the development of boosting and other regression models to predict carbon dioxide emissions for light-duty vehicle designs, with the aim of creating ensemble learning models that leverage vehicle specifications to forecast emissions. Our proposed boosting model is capable of accurately predicting CO2 emissions, even with only one car attribute as input. Moreover, our regression models, in conjunction with the boosting algorithm, can effectively make predictions from various vehicle inputs. Our proposed technique, categorical boosting (Catboost), provides critical insights into transportation-generated air pollution, offering valuable recommendations for both vehicle users and manufacturers. Importantly, Catboost performs data processing in less time and with less memory than other algorithms proposed in the literature. Future research efforts should focus on developing higher performance models and expanding datasets to further improve the accuracy of predictions.

Turbines are core components in jet engines for flight propulsion, power plants, and other important energy conversion processes. They are composed of successive rows of blades so that wakes of upstream blades reach subsequent blades where they perturb the flow in an unsteady manner. At the point where a wake reaches the downstream blades, the perturbation forms a so-called negative jet. In this work, we show that the negative jet partially fulfills the conditions of an anti-splat. Based on this finding, we enhance an anti-splat detection algorithm developed by the present authors in previous work and apply it to direct numerical simulation data of a turbine cascade with unsteady wakes. This provides a sound framework and suitable visualization approaches to investigate the phenomenon even in very complex conditions, as is the alteration of the boundary layer flow along the pressure side of a turbine blade. The approach allows a very clear visualization of this interaction, which was not possible to evidence with previous methods, providing new insight into the physics of this flow. The use of flow paths shows up to which point wakes affect the boundary layer along the blade. The reported physical analysis, made possible by the proposed approach, demonstrates the usefulness of the method for the application domain. The generalization to flows in compressors, pumps, and blade-tower interaction in wind engineering and other fields is possible.

Analogously to the well-known Langevin Monte Carlo method, in this article we provide a method to sample from a target distribution π by simulating a solution of a stochastic differential equation. Hereby, the stochastic differential equation is driven by a general Lévy process which—unlike the case of Langevin Monte Carlo—allows for non-smooth targets. Our method will be fully explored in the particular setting of target distributions supported on the half-line (0 , ∞) and a compound Poisson driving noise. Several illustrative examples conclude the article.

This manuscript discusses a novel estimation approach for parametric hierarchical Archimedean copula. The parameters and structure of this copula are simultaneously estimated while imposing a non-concave penalty on differences between parameters which coincides with an implicit penalty on the copula's structure. The asymptotic properties of the resulting penalized estimator are studied and small sample properties are illustrated using simulations.

In this paper, we compute closed-form expressions of moments and comoments for the CIR process which allows us to provide a new construction of the transition probability density based on a moment argument that differs from the historic approach. For Bates’ model with stochastic volatility and jumps, we show that finite difference approximations of higher moments such as the skewness and the kurtosis are unstable and, as a remedy, provide exact analytic formulas for log-returns. Our approach does not assume a constant mean for log-price differentials but correctly incorporates volatility resulting from Ito’s lemma. We also provide R, MATLAB, and Mathematica modules with exact implementations of the theoretical conditional and unconditional moments. These modules should prove useful for empirical research.

Fragment-based drug discovery begins with the identification of small molecules with a molecular weight of usually less than 250 Da which weakly bind to the protein of interest. This technique is challenging for computational docking methods as binding is determined by only a few specific interactions. Inaccuracies in the energy function or slight deviations in the docking pose can lead to the prediction of incorrect binding or difficulties in ranking fragments in in silico screening. Here, we test RosettaLigand by docking a series of fragments to a cysteine-depleted variant of the TIM-barrel protein, HisF (UniProtKB Q9X0C6). We compare the computational results with experimental NMR spectroscopy screens. NMR spectroscopy gives details on binding affinities of individual ligands, which allows assessment of the ligand-ranking ability using RosettaLigand and also provides feedback on the location of the binding pocket, which serves as a reliable test of RosettaLigand's ability to identify plausible binding poses. From a library screen of 3456 fragments, we identified a set of 31 ligands with intrinsic affinities to HisF with dissociation constants as low as 400 $mu$M. The same library of fragments was blindly screened in silico. RosettaLigand was able to rank binders before non-binders with an area under the curve of the receiver operating characteristics of 0.74. The docking poses observed for binders agreed with the binding pocket identified by NMR chemical shift perturbations for all fragments. Taken together, these results provide a baseline performance of RosettaLigand in a fragment-based drug discovery setting.

Machine learning is applied in a multitude of sectors with very impressive results. This success is due to the availability of an ever-growing amount of data acquired by omnipresent sensor devices and platforms on the internet. But there is a scarcity of labeled data which is required for most ML methods. However, generation of labeled data requires much time and resources. In this paper, we propose a portable, Open Source, simple and responsive manual Tool for 2D multiple object Tracking Annotation (TmoTA). Besides responsiveness, our tool design provides several features like view centering and looped playback that speed up the annotation process. We evaluate our proposed tool by comparing TmoTA with the widely used manual labeling tools CVAT, Label Studio, and two semi-automated tools Supervisely and VATIC with respect to object labeling time and accuracy. The evaluation includes a user study and pre-case studies showing that the annotation time per object frame can be reduced by 20% to 40% over the first 20 annotated objects compared to the manual labeling tools.

Particle methods are a widely used class of algorithms for computer simulation of complex phenomena in various fields, such as fluid dynamics, plasma physics, molecular chemistry, and granular flows, using diverse simulation methods, including Smoothed Particle Hydrodynamics (SPH), Particle-in-Cell (PIC) methods, Molecular Dynamics (MD), and Discrete Element Methods (DEM). Despite the increasing use of particle methods driven by improved computing performance, the relation between these algorithms remains formally unclear, and a unifying formal definition of particle methods is lacking. Here, we present a rigorous mathematical definition of particle methods and demonstrate its importance by applying it to various canonical and non-canonical algorithms, using it to prove a theorem about multi-core parallelizability, and designing a principled scientific computing software based on it. We anticipate that our formal definition will facilitate the solution of complex computational problems and the implementation of understandable and maintainable software frameworks for computer simulation.

These are the full local proceedings of ISSEP 2023. They complement the proceedings published in the Springer LNCS series (vol. 14296, doi:10.1007/978-3-031-44900-0).

In this paper, we present a novel methodology for automatic adaptive weighting of Bayesian Physics-Informed Neural Networks (BPINNs), and we demonstrate that this makes it possible to robustly address multi-objective and multiscale problems. BPINNs are a popular framework for data assimilation, combining the constraints of Uncertainty Quantification (UQ) and Partial Differential Equation (PDE). The relative weights of the BPINN target distribution terms are directly related to the inherent uncertainty in the respective learning tasks. Yet, they are usually manually set a-priori, that can lead to pathological behavior, stability concerns, and to conflicts between tasks which are obstacles that have deterred the use of BPINNs for inverse problems with multiscale dynamics. The present weighting strategy automatically tunes the weights by considering the multitask nature of target posterior distribution. We show that this remedies the failure modes of BPINNs and provides efficient exploration of the optimal Pareto front. This leads to better convergence and stability of BPINN training while reducing sampling bias. The determined weights moreover carry information about task uncertainties, reflecting noise levels in the data and adequacy of the PDE model. We demonstrate this in numerical experiments in Sobolev training, and compare them to analytically ε-optimal baseline, and in a multiscale Lotka-Volterra inverse problem. We eventually apply this framework to an inpainting task and an inverse problem, involving latent field recovery for incompressible flow in complex geometries.

Artificial benchmark functions are commonly used in optimization research because of their ability to rapidly evaluate potential solutions, making them a preferred substitute for real-world problems. However, these benchmark functions have faced criticism for their limited resemblance to real-world problems. In response, recent research has focused on automatically generating new benchmark functions for areas where established test suites are inadequate. These approaches have limitations, such as the difficulty of generating new benchmark functions that exhibit exploratory landscape analysis (ELA) features beyond those of existing benchmarks. The objective of this work is to develop a method for generating benchmark functions for single-objective continuous optimization with user-specified structural properties. Specifically, we aim to demonstrate a proof of concept for a method that uses an ELA feature vector to specify these properties in advance. To achieve this, we begin by generating a random sample of decision space variables and objective values. We then adjust the objective values using CMA-ES until the corresponding features of our new problem match the predefined ELA features within a specified threshold. By iteratively transforming the landscape in this way, we ensure that the resulting function exhibits the desired properties. To create the final function, we use the resulting point cloud as training data for a simple neural network that produces a function exhibiting the target ELA features. We demonstrate the effectiveness of this approach by replicating the existing functions of the well-known BBOB suite and creating new functions with ELA feature values that are not present in BBOB.

Extracting level sets from scalar data is a fundamental operation in visualization with many applications. Recently, the concept of level set extraction has been extended to bivariate scalar fields. Prior work on vector field equivalence, wherein an analyst marks a region in the domain and is shown other regions in the domain with similar vector values, pointed out the need to make this extraction operation fast, so that analysts can work interactively. To date, the fast extraction of level sets from bivariate scalar fields has not been researched as extensively as for the univariate case. In this paper, we present a novel algorithm that extracts fiber lines, i.e., the preimages of so called control polygons (FSCP), for bivariate 2D data by joint traversal of bounding volume hierarchies for both grid and FSCP elements. We performed an extensive evaluation, comparing our method to a two-dimensional adaptation of the method proposed by Klacansky et al., as well as to the naive approach for fiber line extraction. The evaluation incorporates a vast array of configurations in several datasets. We found that our method provides a speedup of several orders of magnitudes compared to the naive algorithm and requires two thirds of the computation time compared to Klacansky et al. adapted for 2D.

The Archive Query Log (AQL) is a previously unused, comprehensive query log collected at the Internet Archive over the last 25 years. Its first version includes 356 million queries, 166 million search result pages, and 1.7 billion search results across 550 search providers. Although many query logs have been studied in the literature, the search providers that own them generally do not publish their logs to protect user privacy and vital business data. Of the few query logs publicly available, none combines size, scope, and diversity. The AQL is the first to do so, enabling research on new retrieval models and (diachronic) search engine analyses. Provided in a privacy-preserving manner, it promotes open research as well as more transparency and accountability in the search industry.

Privacy-preserving record linkage (PPRL) determines records representing the same entitywhile guaranteeing the privacy of individuals. A common approach is to encode plaintext data ofrecords into Bloom filters that enable efficient calculation of similarities. A crucial step of PPRL isthe classification of Bloom filter pairs as match or non-match based on computed similarities. In thecontext of record linkage, several weighting schemes and classification methods are available. Themajority of weighting methods determine and adapt weights by applying the Fellegi&Sunter modelfor each attribute. In the PPRL domain, the attributes of a record are encoded in a joint record-levelBloom filter to impede cryptanalysis attacks so that the application of existing attribute-wise weightingapproaches is not feasible. We study methods that use attribute-specific weights in record-levelencodings and integrate weight adaptation approaches based on individual value frequencies. Theexperiments on real-world datasets show that frequency-dependent weighting schemes improve thelinkage quality as well as the robustness with regard to the threshold selection.

Graph metrics, such as the simple but popular vertex degree and others based on it, are well defined for static graphs. However, adapting static metrics for temporal graphs is still part of current research. In this paper, we propose a set of temporal extensions of four degree-dependent metrics, as well as aggregations like minimum, maximum, and average degree of (i) a vertex over a time interval and (ii) a graph at a specific point in time. We show why using the static degree can lead to wrong assumptions about the relevance of a vertex in a temporal graph and highlight the need to include time as a dimension in the metric. We propose a baseline algorithm to calculate the degree evolution of all vertices in a temporal graph and show its implementation in a distributed in-memory dataflow system. Using real-world and synthetic datasets containing up to 462 million vertices and 1.7 billion edges, we show the scalability of our algorithm on a distributed cluster achieving a speedup of around 12 on 16 machines.

Stylized tubes are an established visualization primitive for line data as encountered in many scientific fields, ranging from characteristic lines in flow fields, fiber tracks reconstructed from diffusion tensor imaging, to trajectories of moving objects as they arise from cyber-physical systems in many engineering disciplines. Typical challenges include large data set sizes demanding for efficient rendering techniques as well as a large number of attributes that cannot be mapped simultaneously to the basic visual attributes provided by a tube-based visualization. In this work, we tackle both challenges with a new on-tube visualization approach. We improve recent work on high-quality GPU ray casting of Hermite spline tubes supporting ambient occlusion and extend it by a new layered procedural texturing technique. In the proposed framework, a large number of data set attributes can be mapped simultaneously to a variety of glyphs and plots that are embedded in texture space and organized in layers. Efficient rendering with minimal data transfer is achieved by generating the glyphs procedurally and drawing them in a deferred shading pass. We integrated these techniques in a prototype visualization tool that facilitates flexible mapping of data set attributes to visual tube and glyph attributes. We studied our approach on a variety of example data from different fields and found it to provide a highly adaptable and extensible toolbox to quickly craft tailor-made tube-based trajectory visualizations.

The use of deep machine learning (ML) in protein structure prediction has made it possible to easily access a large number of annotated conformations that can potentially compensate for missing experimental structures in structure-based drug discovery (SBDD). However, it is still unclear whether the accuracy of these predicted conformations is sufficient for screening chemical compounds that will effectively interact with a protein target for pharmacological purposes. In this opinion article, we examine the potential benefits and limitations of using state-annotated conformations for ultra-large library screening (ULLS) in light of the growing size of ultra-large libraries (ULLs). We believe that targeting different conformational states of common drug targets like G-protein-coupled receptors (GPCRs), which can regulate human physiology by switching between different conformations, can offer multiple advantages.

Hybrid User Interfaces (HUIs) describe a combination of several devices into one system. Such a combination promises to alleviate the disadvantages of one device by adding features of another. The term HUIs was first coined by Feiner and Shamash [10] in 1991. Since then, many research projects have focused on exploring possible systems within this interface category, showing the importance and interest of this term. However, the broadness of the follow-up research projects can be seen as "wild growth", which caused the term HUI to become unspecific and fragmented - as also the call for submission of this workshop states [14]. Additionally, a set of related terms have been used since then, which partially overlap with HUI, like cross-device interfaces or augmented displays.

Software projects are complex technical and organizational systems involving large numbers of artifacts and developers. To understand and tame software complexity, a wide variety of program analysis techniques have been developed for bug detection, program comprehension, verification, and more. At the same time, repository mining techniques aim at obtaining insights into the inner socio-technical workings of software projects at a larger scale. While both program analysis and repository mining have been successful on their own, they are largely isolated, which leaves considerable potential for synergies untapped. We present SEAL, the first integrated approach that combines low-level program analysis with high-level repository information. SEAL maps repository information, mined from the development history of a project, onto a low-level intermediate program representation, making it available for state-of-the-art program analysis. SEAL's integrated approach allows us to efficiently address software engineering problems that span multiple levels of abstraction, from low-level data flow to high-level organizational information. To demonstrate its merits and practicality, we use SEAL to determine which code changes modify central parts of a given software project, how authors interact (indirectly) with each other through code, and we demonstrate that putting static analysis' results into a socio-technical context improves their expressiveness and interpretability.

The design and choice of benchmark suites are ongoing topics of discussion in the multi-objective optimization community. Some suites provide a good understanding of their Pareto sets and fronts, such as the well-known DTLZ and ZDT problems. However, they lack diversity in their landscape properties and do not provide a mechanism for creating multiple distinct problem instances. Other suites, like bi-objective BBOB, possess diverse and challenging landscape properties, but their optima are not well understood and can only be approximated empirically without any guarantees. This work proposes a methodology for creating complex continuous problem landscapes by concatenating single-objective functions from version 2 of the multiple peaks model (MPM2) generator. For the resulting problems, we can determine the distribution of optimal points with arbitrary precision w.r.t. a measure such as the dominated hypervolume. We show how the properties of the MPM2 generator influence the multi-objective problem landscapes and present an experimental proof-of-concept study demonstrating how our approach can drive well-founded benchmarking of MO algorithms.

Evolutionary scenarios describing the evolution of a family of genes within a collection of species comprise the mapping of the vertices of a gene tree T to vertices and edges of a species tree S. The relative timing of the last common ancestors of two extant genes (leaves of T) and the last common ancestors of the two species (leaves of S) in which they reside is indicative of horizontal gene transfers (HGT) and ancient duplications. Orthologous gene pairs, on the other hand, require that their last common ancestors coincides with a corresponding speciation event. The relative timing information of gene and species divergences is captured by three colored graphs that have the extant genes as vertices and the species in which the genes are found as vertex colors: the equal-divergence-time (EDT) graph, the later-divergence-time (LDT) graph and the prior-divergence-time (PDT) graph, which together form an edge partition of the complete graph.

The placenta remains the key organ to pregnancy complications, such as preeclampsia, contrarily the pathophysiology underlying the placental dysfunctions remains elusive. Here, we present our Disease Map ``NaviCenta'', which is an online resource based on the interactions between tissues, cellular compartments, and molecules that mediate disease-related processes in the placenta. We built cellular and molecular interaction networks based upon manual curation and annotation of publicly available information in the scientific literature, pathways resources, and Omics data. NaviCenta (Navigate the plaCenta) serves as an open access, spatio-temporal, multi-scale knowledge base, and analytical tool for enhanced interpretation and hypothesis testing on various placental disease phenotypes.

Orientation and finding ways in buildings are particular challenges for people with disabilities. To cope with these obstacles, open accessible indoor maps enriched with accessibility information are required. To generate these is both a time-consuming and knowledge-intensive task [6] One possible solution is the use of volunteers who collect freely accessible map data. To support so-called mappers in this process, appropriate user-centered applications are needed, that guide and motivate those who collect them in the acquisition. Therefore, we present the results and their conclusions from a study carried out in the German-speaking area, which was primarily aimed at active mappers from the OpenStreetMap community.

In diesem zweigeteilten Workshop werden zuerst grundlegende Kompetenzen im Bereich Künstliche Intelligenz aufgebaut, Lehr-Lern-Materialien erkundet und anschließend gemeinsam mit Lehrkräften und Forschenden über die Vermittlung von KI im Informatikunterricht diskutiert.

We introduce small-text, an easy-to-use active learning library, which offers pool-based active learning for single- and multi-label text classification in Python. It features numerous pre-implemented state-of-the-art query strategies, including some that leverage the GPU. Standardized interfaces allow the combination of a variety of classifiers, query strategies, and stopping criteria, facilitating a quick mix and match, and enabling a rapid development of both active learning experiments and applications. With the objective of making various classifiers and query strategies accessible for active learning, small-text integrates several well-known machine learning libraries, namely scikit-learn, Pytorch, and Hugging Face transformers. The latter integrations are optionally installable extensions, so GPUs can be used but are not required. Using this new library, we investigate the performance of the recently published SetFit training paradigm, which we compare to vanilla transformer fine-tuning, finding that it matches the latter in classification accuracy while outperforming it in area under the curve. The library is available under the MIT License at urlhttps://github.com/webis-de/small-text, in version 1.3.0 at the time of writing.

We present a higher-order convergent numerical solver for active polar hydrodynamics in three-dimensional domains of arbitrary shape, along with a scalable open-source software implementation for shared- and distributed-memory parallel computers. This enables the computational study of the nonlinear dynamics of out-of-equilibrium materials from first principles. We numerically solve the nonlinear active Ericksen-Leslie hydrodynamic equations of three-dimensional (3D) active nematics using both a meshfree and a hybrid particle-mesh method in either the Eulerian or Lagrangian frame of reference. The solver is validated against a newly derived analytical solution in 3D and implemented using the OpenFPM software library for scalable scientific computing. We then apply the presented method to studying the transition of 3D active polar fluids to spatiotemporal chaos, the emergence of coherent angular motion in a 3D annulus, and chiral vortices in symmetric and asymmetric 3D shapes resembling dividing cells. Overall, this provides a robust and efficient open-source simulation framework for 3D active matter with verified numerical convergence and scalability on parallel computers.

Active polar fluids exhibit spontaneous flow when sufficient active stress is generated by internal molecular mechanisms. This is also referred to as an active Fréedericksz transition. Experiments have revealed the existence of competing in-plane and out-of-plane instabilities in three-dimensional active matter. So far, however, a theoretical model reconciling all observations is missing. In particular, the role of boundary conditions in these instabilities still needs to be explained. Here, we characterize the spontaneous flow transition in a symmetry-preserving three-dimensional active Ericksen-Leslie model, showing that the boundary conditions select the emergent behavior. Using nonlinear numerical solutions and linear perturbation analysis, we explain the mechanism for both in-plane and out-of-plane instabilities under extensile active stress for perpendicular polarity anchoring at the boundary, whereas parallel anchoring only permits in-plane flows under contractile stress or out-of-plane wrinkling under extensile stress.

The wind is a crucial factor in various domains such as weather forecasting, the wind power industry, agriculture, structural health monitoring, and so on. The variability and unpredictable nature of the wind is a challenge faced by most wind-energy-based sectors. Several atmospheric and geographical factors influence wind characteristics. Many wind forecasting methods and tools have been introduced since early times. Wind forecasting can be carried out short-, medium-, and long-term. The uncertainty factors of the wind challenge the accuracy of techniques. This article brings the general background of physical, statistical, and intelligent approaches and their methods used to predict wind characteristics and their challenges---this work's objective is to improve effective data-driven models for forecasting wind-power production. The investigation and listing of the effectiveness of improved machine learning models to estimate univariate wind-energy time-based data is crucially the prominent focus of this work. The performance of various ML predicting models was examined using ensemble learning (ES) models, such as boosted trees and bagged trees, Support Vector Regression (SVR) with distinctive kernels etc. Numerous neural networks have recently been constructed for forecasting wind speed and power due to artificial intelligence (AI) advancement. Based on the model summary, further directions for research and application developments can be planned.

The higher-order correlation clustering problem is an expressive model, and recently, local search heuristics have been proposed for several applications. Certifying optimality, however, is NP-hard and practically hampered already by the complexity of the problem statement. Here, we focus on establishing partial optimality conditions for the special case of complete graphs and cubic objective functions. In addition, we define and implement algorithms for testing these conditions and examine their effect numerically, on two datasets.

Cytotoxic cancer therapy often results in dose-limiting haematotoxic side effects. Predicting an individual's risk is a major objective in precision medicine of cancer treatment. In this regard, patient heterogeneity presents a significant challenge. In this paper, we explore the use of hypothesis-free machine learning models based on recurrent nonlinear auto-regressive networks with exogenous inputs (NARX) as an approach to achieve this goal. Also, we propose a knowledge transfer approach to ameliorate the issue of sparse individual data, which typically hampers learning of individual networks. We demonstrate the feasibility of our approach based on a virtual patient population generated using a semi-mechanistic model of haematopoiesis and imposing different cytotoxic therapy scenarios on it. Employing different techniques of model optimisation, we derive robust and parsimonious individual networks with good generalisation performances. Moreover, we analyse in detail possible factors influencing the generalisation performance. Results suggest that our transfer learning approach using NARX networks can provide robust predictions of individual patient's response to treatment. As a practical perspective, we apply our approach to individual time series data of two patients with non-Hodgkin's lymphoma.

With the prevalence of mental disorders such as anxiety rising sharply, new low-threshold and scalable solutions for mental health support are needed - especially for the younger population. In recent years studies investigating the use of apps and chatbots to deliver content based in cognitive behavioral therapy (CBT) for mental health support showed promising results. The use of voice assistants (VAs) to convey methods and interventions stemming from CBT for students with test anxiety has, however, thus far not been investigated. To address this problem, we present the design and prototypical implementation of a VA for CBT-based interventions against exam anxiety in students. In a user study with the primary target group of students (N=20), it could be shown that the presented system has a good usability and a high level of acceptance (CSQi mean 27.15).

Drought is a natural hazard with complex socioeconomic impacts and influences on human experiences. Preparedness is the only way a society can mitigate drought impacts but integrating participatory decision-making with hydrological modeling into a more comprehensive planning process is still a challenge. Here, we present a step-by-step methodology to guide the implementation of a participatory drought preparedness plan (DPP), specially designed for hydrosystems and cities scales. We highlight strategies to engage local stakeholders in constructing such plans and build trust in the process. We propose two types of drought preparedness plans: (1) Socio-technical–built only from the tacit knowledge of the system operators, which needs only two days to be ready; and (2) Socio-technical with modeling-intensive simulation—a more robust methodology that adds hydrologic and hydraulic modeling to the existing tacit knowledge. The participatory DPP methodology was developed and applied to hydrosystems and cities in a drought-prone area of Brazil. Our findings suggest that modeling was important, but not essential to assessing vulnerability scenarios and strategies. However, the simplified version can achieve satisfactory results even when data and resources are limited. We present the methodology as a nine-step participatory planning methodology developing a meaningful and convincing narrative that speaks to theory and practice.

We provide an internal validation study of a recently published precise DNA mixture algorithm based on Hamiltonian Monte Carlo sampling (Susik et al., 2022). We provide results for all 428 mixtures analysed by Riman et al. (2021) and compare the results with two state-of-the-art software products: STRmix� v2.6 and Euroformix v3.4.0. The comparison shows that the Hamiltonian Monte Carlo method provides reliable values of likelihood ratios (LRs) close to the other methods. We further propose a novel large-scale precision benchmark and quantify the precision of the Hamiltonian Monte Carlo method, indicating its improvements over existing solutions. Finally, we analyse the influence of the factors discussed by Buckleton et al. (2022).

Current approaches to automatic summarization of scientific papers generate informative summaries in the form of abstracts. However, abstracts are not intended to show the relationship between a paper and the references cited in it. We propose a new contextualized summarization approach that can generate an informative summary conditioned on a given sentence containing the citation of a reference (a so-called ``citance''). This summary outlines content of the cited paper relevant to the citation location. Thus, our approach extracts and models the citances of a paper, retrieves relevant passages from cited papers, and generates abstractive summaries tailored to each citance. We evaluate our approach using **Webis-Context-SciSumm-2023**, a new dataset containing 540K computer science papers and 4.6M citances therein.

In a project called Ä Library of a Billion Words" we needed an implementation of the CTS protocol that is capable of handling a text collection containing at least 1 billion words. Because the existing solutions did not work for this scale or were still in development I started an implementation of the CTS protocol using methods that MySQL provides. Last year we published a paper that introduced a prototype with the core functionalities without being compliant with the specifications of CTS (Tiepmar et al., 2013). The purpose of this paper is to describe and evaluate the MySQL based implementation now that it is fulfilling the specifications version 5.0 rc.1 and mark it as finished and ready to use. Further information, online instances of CTS for all described datasets and binaries can be accessed via the projects website.

BACKGROUND: Recent studies have shown that hyperspectral imaging (HSI) combined with neural networks can detect colorectal cancer. Usually, different pre-processing techniques (e.g., wavelength selection and scaling, smoothing, denoising) are analyzed in detail to achieve a well-trained network. The impact of post-processing was studied less. METHODS: We tested the following methods: (1) Two pre-processing techniques (Standardization and Normalization), with (2) Two 3D-CNN models: Inception-based and RemoteSensing (RS)-based, with (3) Two post-processing algorithms based on median filter: one applies a median filter to a raw predictions map, the other applies the filter to the predictions map after adopting a discrimination threshold. These approaches were evaluated on a dataset that contains ex vivo hyperspectral (HS) colorectal cancer records of 56 patients. RESULTS: (1) Inception-based models perform better than RS-based, with the best results being 92% sensitivity and 94% specificity; (2) Inception-based models perform better with Normalization, RS-based with Standardization; (3) Our outcomes show that the post-processing step improves sensitivity and specificity by 6.6% in total. It was also found that both post-processing algorithms have the same effect, and this behavior was explained. CONCLUSION: HSI combined with tissue classification algorithms is a promising diagnostic approach whose performance can be additionally improved by the application of the right combination of pre- and post-processing.

Machine learning drives forward the development in many areas of Natural Language Processing (NLP). Until now, many NLP systems and research are focusing on high-resource languages, i.e. languages for which many data resources exist. Recently, so-called low-resource languages increasingly come into focus. In this context, multi-lingual language models, which are trained on related languages to a target low-resource language, may enable NLP tasks on this low-resource language. In this work, we investigate the use of multi-lingual models for Named Entity Recognition (NER) for low-resource languages. We consider the West Slavic language family and the low-resource languages Upper Sorbian and Kashubian. Three RoBERTa models were trained from scratch, two mono-lingual models for Czech and Polish, and one bi-lingual model for Czech and Polish. These models were evaluated on the NER downstream task for Czech, Polish, Upper Sorbian, and Kashubian, and compared to existing state-of-the-art models such as RobeCzech, HerBERT, and XLM-R. The results indicate that the mono-lingual models perform better on the language they were trained on, and both the mono-lingual and language family models outperform the large multi-lingual model in downstream tasks. Overall, the study shows that low-resource West Slavic languages can benefit from closely related languages and their models.

Assumption-based Argumentation (ABA) is a well-known structured argumentation formalism, whereby arguments and attacks between them are drawn from rules, defeasible assumptions and their contraries. A common restriction imposed on ABA frameworks (ABAFs) is that they are flat, i.e., each of the defeasible assumptions can only be assumed, but not derived. While it is known that flat ABAFs can be translated into abstract argumentation frameworks (AFs) as proposed by Dung, no translation exists from general, possibly non-flat ABAFs into any kind of abstract argumentation formalism. In this paper, we close this gap and show that bipolar AFs (BAFs) can instantiate general ABAFs. To this end we develop suitable, novel BAF semantics which borrow from the notion of deductive support. We investigate basic properties of our BAFs, including computational complexity, and prove the desired relation to ABAFs under several semantics. Finally, in order to support computation and explainability, we propose the notion of dispute trees for our BAF semantics.

AbstractDuring the RESOLVE project (``High‐resolution imaging in subsurface geophysics: development of a multi‐instrument platform for interdisciplinary research''), continuous surface displacement and seismic array observations were obtained on Glacier d'Argentière in the French Alps for 35 days in May 2018. The data set is used to perform a detailed study of targeted processes within the highly dynamic cryospheric environment. In particular, the physical processes controlling glacial basal motion are poorly understood and remain challenging to observe directly. Especially in the Alpine region for temperate based glaciers where the ice rapidly responds to changing climatic conditions and thus, processes are strongly intermittent in time and heterogeneous in space. Spatially dense seismic and Global Positioning System (GPS) measurements are analyzed applying machine learning to gain insight into the processes controlling glacial motions of Glacier d'Argentière. Using multiple bandpass‐filtered copies of the continuous seismic waveforms, we compute energy‐based features, develop a matched field beamforming catalog and include meteorological observations. Features describing the data are analyzed with a gradient boosting decision tree model to directly estimate the GPS displacements from the seismic noise. We posit that features of the seismic noise provide direct access to the dominant parameters that drive displacement on the highly variable and unsteady surface of the glacier. The machine learning model infers daily fluctuations and longer term trends. The results show on‐ice displacement rates are strongly modulated by activity at the base of the glacier. The techniques presented provide a new approach to study glacial basal sliding and discover its full complexity.

Beim selektiven Laserschmelzen wird Metallpulver schichtweise aufgeschmolzen und mit dem bereits gefertigten Bauteil verschmolzen. Innerhalb dieses Prozesses entstehen vermeidbare fehlerhafte Schichten. Derartige Fehler sind erst durch verschiedene Druck- und Zugfestigkeitsexperimente nach dem Druck feststellbar. Dieses Vorgehen ist kostenintensiv und ineffizient.

Monitoring the status of large computing systems is essential to identify unexpected behavior and improve their performance and uptime. However, due to the large-scale and distributed design of such computing systems as well as a large number of monitoring parameters, automated monitoring methods should be applied. Such automatic monitoring methods should also have the ability to adapt themselves to the continuous changes in the computing system. In addition, they should be able to identify behavioral anomalies in useful time, to perform appropriate reactions. This work proposes a general lightweight and unsupervised method for near real-time anomaly detection using operational data measurement on large computing systems. The proposed model requires as little as 4 hours of data and 50 epochs for each training process to accurately resemble the behavioral pattern of computing systems.

BACKGROUND: The immune response is a crucial factor for mediating the benefit of cardiac cell therapies. Our previous research showed that cardiomyocyte transplantation alters the cardiac immune response and, when combined with short-term pharmacological CCR2 inhibition, resulted in diminished functional benefit. However, the specific role of innate immune cells, especially CCR2 macrophages on the outcome of cardiomyocyte transplantation, is unclear.METHODS: We compared the cellular, molecular, and functional outcome following cardiomyocyte transplantation in wildtype and T cell- and B cell-deficient Rag2del mice. The cardiac inflammatory response was assessed using flow cytometry. Gene expression profile was assessed using single-cell and bulk RNA sequencing. Cardiac function and morphology were determined using magnetic resonance tomography and immunohistochemistry respectively.RESULTS: Compared to wildtype mice, Rag2del mice show an increased innate immune response at steady state and disparate macrophage response after MI. Subsequent single-cell analyses after MI showed differences in macrophage development and a lower prevalence of CCR2 expressing macrophages. Cardiomyocyte transplantation increased NK cells and monocytes, while reducing CCR2-MHC-IIlo macrophages. Consequently, it led to increased mRNA levels of genes involved in extracellular remodelling, poor graft survival, and no functional improvement. Using machine learning-based feature selection, Mfge8 and Ccl7 were identified as the primary targets underlying these effects in the heart.CONCLUSIONS: Our results demonstrate that the improved functional outcome following cardiomyocyte transplantation is dependent on a specific CCR2 macrophage response. This work highlights the need to study the role of the immune response for cardiomyocyte cell therapy for successful clinical translation.

py-spinnaker2 provides a light-weight Python interface for running experiments on the SpiNNaker2 neuromorphic chip. For low-level users, the so-called ``experiment runner'' allows to define and run reproducible batch experiments on single SpiNNaker2 chips. An experiment is comprised of the ``mem-files'' for the SpiNNaker2 ARM cores (i.e. the instruction code), input data to be sent to specific memory locations, and memory regions to be read after the experiment. High-level users can define spiking neural networks (SNN) and hybrid DNN/SNN models in py-spinnaker2. The networks are automatically partitioned and mapped to a SpiNNaker2 chip, the SNN simulation is executed via the experiment runner and results are provided back to the user in the Python interface. The API for defining SNN is inspired by PyNN (http://neuralensemble.org/PyNN/). If no SpiNNaker2 chip is available, SNN models can also be executed using the brian2 simulator (https://briansimulator.org/).

The term Procedural Content Generation (PCG) refers to the (semi-)automatic generation of game content by algorithmic means, and its methods are becoming increasingly popular in game-oriented research and industry. A special class of these methods, which is commonly known as search-based PCG, treats the given task as an optimisation problem. Such problems are predominantly tackled by evolutionary algorithms. We will demonstrate in this paper that obtaining more information about the defined optimisation problem can substantially improve our understanding of how to approach the generation of content. To do so, we present and discuss three efficient analysis tools, namely diagonal walks, the estimation of high-level properties, as well as problem similarity measures. We discuss the purpose of each of the considered methods in the context of PCG and provide guidelines for the interpretation of the results received. This way we aim to provide methods for the comparison of PCG approaches and eventually, increase the quality and practicality of generated content in industry.

Wikidata, now a decade old, is the largest public knowledge graph, with data on more than 100 million concepts contributed by over 560,000 editors. It is widely used in applications and research. At its launch in late 2012, however, it was little more than a hopeful new Wikimedia project, with no content, almost no community, and a severely restricted platform. Seven years earlier still, in 2005, it was merely a rough idea of a few PhD students, a conceptual nucleus that had yet to pick up many important influences from others to turn into what is now called Wikidata. In this paper, we try to recount this remarkable journey, and we review what has been accomplished, what has been given up on, and what is yet left to do for the future.

Nystagmus describes an involuntary oscillation of one or both eyes. Depending on its origin, it decreases visual acuity and may lead to vertigo, oscillopsia, physical strain, and social anxiety. Available assistive technology only accounts for the limited sight, rather than compensating the nystagmus itself. When designing and using a system for nystagmus compensation, it is necessary to create a user model that covers the usertextquoterights nystagmus as detailed as possible, due to the various specific manifestations of the illness. Determining nystagmus parameters involves a time-consuming manual inspection of eye movement recordings and requires considerable expertise. So far, no algorithm for automatic nystagmus parameter detection and categorization of eye movements has been established. Therefore, the here presented paper strives to address this gap by presenting a novel approach for automatic nystagmus parameter determination and real-time eye movement categorization. The algorithm for automatically determining nystagmus parameters is evaluated through a pilot study with subjects from the target group, focusing on the accuracy of parameter detection.

This paper proposes a spatial–temporal recurrent neural network architecture for deep Q-networks that can be used to steer an autonomous ship. The network design makes it possible to handle an arbitrary number of surrounding target ships while offering robustness to partial observability. Furthermore, a state-of-the-art collision risk metric is proposed to enable an easier assessment of different situations by the agent. The COLREG rules of maritime traffic are explicitly considered in the design of the reward function. The final policy is validated on a custom set of newly created single-ship encounters called textquoteleftAround the Clocktextquoteright problems and the commonly used Imazu (1987) problems, which include 18 multi-ship scenarios. Performance comparisons with artificial potential field and velocity obstacle methods demonstrate the potential of the proposed approach for maritime path planning. Furthermore, the new architecture exhibits robustness when it is deployed in multi-agent scenarios and it is compatible with other deep reinforcement learning algorithms, including actor-critic frameworks.

This paper proposes a realistic modularized framework for controlling autonomous surface vehicles (ASVs) on inland waterways (IWs) based on deep reinforcement learning (DRL). The framework improves operational safety and comprises two levels: a high-level local path planning (LPP) unit and a low-level path following (PF) unit, each consisting of a DRL agent. The LPP agent is responsible for planning a path under consideration of dynamic vessels, closing a gap in the current research landscape. In addition, the LPP agent adequately considers traffic rules and the geometry of the waterway. We thereby introduce a novel application of a spatial-temporal recurrent neural network architecture to continuous action spaces. The LPP agent outperforms a state-of-the-art artificial potential field (APF) method by increasing the minimum distance to other vessels by 65% on average. The PF agent performs low-level actuator control while accounting for shallow water influences and the environmental forces winds, waves, and currents. Compared with a proportional-integral-derivative (PID) controller, the PF agent yields only 61% of the mean cross-track error (MCTE) while significantly reducing control effort (CE) in terms of the required absolute rudder angle. Lastly, both agents are jointly validated in simulation, employing the lower Elbe in northern Germany as an example case and using real automatic identification system (AIS) trajectories to model the behavior of other ships.

This paper has two objectives. Firstly, to introduce a new framework XF -OPTIMETA for testing and comparing Hyperparameter Optimization (HPO) methods. The framework supports model-free methods, e.g., Random Search (RS), as well as model-based methods, such as Bayesian Optimization (BO), with various surrogate models. Due to the generalized and modular structure of the XF-OPTIMETA framework, it can be easily extended to other optimization methods for dif-ferent optimization problems. The second objective is to empir-ically compare the performance of various HPO methods for population-based metaheuristics. For that the XF -OPTIMETA framework is used to apply HPO methods to the Hierarchical Simple Probabilistic Population-Based Optimization (H-SPPBO) metaheuristic for the Dynamic Traveling Salesperson Problem (DTSP) and to calculate high-performing parameter values for H-SPPBO. Promising results are obtained using the parameter values found by BO. In particular, a parameter set obtained with Gradient-Boosted Regression Trees (GBRT) outperforms a reference parameter set for H-SPPBO from an existing study.

Context awareness and adaption are an important requirement for data streaming applications in the Internet of Things (IoT). Concept drift arises under changing context and many solutions for drift detection and system adaption have been developed in the past. Even though the computing resources of these approaches are not negligible, performance aspects, e.g. runtime or memory usage and scalability have not been investigated adequately. The goal of this thesis is to fill this gap and to provide a mean for performance investigations of concept drift handling methods and other approaches that are developed in the research field Data Science. Therefore performance benchmarking of state-of-the-art solutions is conducted. Moreover the thesis discusses performance bottlenecks of the approaches and demonstrates possible improvements based on the knowledge gained by leveraging tools and methods from the performance analysis domain. Approaches are then implemented into an IoT application and deployed with stream processing engines for a final evaluation.

Concept drift detection is crucial for many AI systems to ensure the system's reliability. These systems often have to deal with large amounts of data or react in real time. Thus, drift detectors must meet computational requirements or constraints with a comprehensive performance evaluation. However, so far, the focus of developing drift detectors is on detection quality, e.g.~accuracy, but not on computational performance, such as running time. We show that the previous works consider computational performance only as a secondary objective and do not have a benchmark for such evaluation. Hence, we propose a set of metrics that considers both, computational performance and detection quality. Among others, our set of metrics includes the Relative Runtime Overhead RRO to evaluate a drift detector's computational impact on an AI system. This work focuses on unsupervised drift detectors, not being restricted to the availability of labeled data. We measure the computational performance based on the RRO and memory consumption of four available unsupervised drift detectors on five different data sets. The range of the RRO reaches from 1.01 to 20.15. Moreover, we measure state-of-the-art detection quality metrics to discuss our evaluation results and show the necessity of thorough computational performance considerations for drift detectors. Additionally, we highlight and explain requirements for a comprehensive benchmark of drift detectors. Our investigations can also be extended for supervised drift detection.

Simulating abundances of stable water isotopologues, that is, molecules differing in their isotopic composition, within climate models allows for comparisons with proxy data and, thus, for testing hypotheses about past climate and validating climate models under varying climatic conditions. However, many models are run without explicitly simulating water isotopologues. We investigate the possibility of replacing the explicit physics-based simulation of oxygen isotopic composition in precipitation using machine learning methods. These methods estimate isotopic composition at each time step for given fields of surface temperature and precipitation amount. We implement convolutional neural networks (CNNs) based on the successful UNet architecture and test whether a spherical network architecture outperforms the naive approach of treating Earth’s latitude-longitude grid as a flat image. Conducting a case study on …

The vast and heterogeneous data being constantly generated in clinics can provide great wealth for patients and research alike. The quickly evolving field of medical informatics research has contributed numerous concepts, algorithms, and standards to facilitate this development. However, these difficult relationships, complex terminologies, and multiple implementations can present obstacles for people who want to get active in the field. With a particular focus on medical informatics research conducted in Germany, we present in our Viewpoint a set of 10 important topics to improve the overall interdisciplinary communication between different stakeholders (eg, physicians, computational experts, experimentalists, students, patient representatives). This may lower the barriers to entry and offer a starting point for collaborations at different levels. The suggested topics are briefly introduced, then general best practice guidance is given, and further resources for in-depth reading or hands-on tutorials are recommended. In addition, the topics are set to cover current aspects and open research gaps of the medical informatics domain, including data regulations and concepts; data harmonization and processing; and data evaluation, visualization, and dissemination. In addition, we give an example on how these topics can be integrated in a medical informatics curriculum for higher education. By recognizing these topics, readers will be able to (1) set clinical and research data into the context of medical informatics, understanding what is possible to achieve with data or how data should be handled in terms of data privacy and storage; (2) distinguish current interoperability standards and obtain first insights into the processes leading to effective data transfer and analysis; and (3) value the use of newly developed technical approaches to utilize the full potential of clinical data.

RNA features a highly negatively charged phosphate backbone that attracts a cloud of counter-ions that reduce the electrostatic repulsion in a concentration dependent manner. Ion concentrations thus have a large influence on folding and stability of RNA structures. Despite their well-documented effects, salt effects are not handled consistently by currently available secondary structure prediction algorithms. Combining Debye-Hückel potentials for line charges and Manning’s counter-ion condensation theory, Einert et al. (Biophys J 100: 2745-2753, 2011) modeled the energetic contributions of monovalent cations on loops and helices.

Biofilm-forming benthic diatoms are key primary producers in coastal habitats, where they frequently dominate sunlit intertidal substrata. The development of gliding motility in raphid diatoms was a key molecular adaptation that contributed to their evolutionary success. However, the structure-function correlation between diatom adhesives utilized for gliding and their relationship to the extracellular matrix that constitutes the diatom biofilm is unknown. Here, we have used proteomics, immunolocalization, comparative genomics, phylogenetics and structural homology analysis to investigate the evolutionary history and function of diatom adhesive proteins. Our study identified eight proteins from the adhesive trails of Craspedostauros australis, of which four form a new protein family called Trailins that contain an enigmatic Choice-of-Anchor A (CAA) domain, which was acquired through horizontal gene transfer from bacteria. Notably, the CAA-domain shares a striking structural similarity with one of the most widespread domains found in ice-binding proteins (IPR021884). Our work offers new insights into the molecular basis for diatom biofilm formation, shedding light on the function and evolution of diatom adhesive proteins. This discovery suggests that there is a transition in the composition of biomolecules required for initial surface colonization and those utilized for 3D biofilm matrix formation.

Biofilm-forming benthic diatoms are key primary producers in coastal habitats, where they frequently dominate sunlit intertidal substrata. The development of gliding motility in raphid diatoms was a key molecular adaptation that contributed to their evolutionary success. However, the structure-function correlation between diatom adhesives utilized for gliding and their relationship to the extracellular matrix that constitutes the diatom biofilm is unknown. Here, we have used proteomics, immunolocalization, comparative genomics, phylogenetics and structural homology analysis to investigate the evolutionary history and function of diatom adhesive proteins. Our study identified eight proteins from the adhesive trails of Craspedostauros australis, of which four form a new protein family called Trailins that contain an enigmatic Choice-of-Anchor A (CAA) domain, which was acquired through horizontal gene transfer from bacteria. Notably, the CAA-domain shares a striking structural similarity with one of the most widespread domains found in ice-binding proteins (IPR021884). Our work offers new insights into the molecular basis for diatom biofilm formation, shedding light on the function and evolution of diatom adhesive proteins. This discovery suggests that there is a transition in the composition of biomolecules required for initial surface colonization and those utilized for 3D biofilm matrix formation.

This paper evaluates the viability of using fixed language models for training text classification networks on low-end hardware. We combine language models with a CNN architecture and put together a comprehensive benchmark with 8 datasets covering single-label and multi-label classification of topic, sentiment, and genre. Our observations are distilled into a list of trade-offs, concluding that there are scenarios, where not fine-tuning a language model yields competitive effectiveness at faster training, requiring only a quarter of the memory compared to fine-tuning.

Given the assumption that weather risks affect crop yields, we designed a weather index insurance product for soybean producers in the US state of Illinois. By separating the entire vegetation cycle into four growth stages, we investigate whether the phase-division procedure contributes to weather-yield loss relation estimation and, hence, to basis risk mitigation. Concretely, supposing stage-variant interaction patterns between temperature-based weather index growing degree days and rainfall-based weather index cumulative rainfall, a nonparametric weather-yield loss relation is estimated by a generalized additive model. The model includes penalized B-spline (P-spline) approach based on nonlinear optimal indemnity solutions under the expected utility framework. The P-spline analysis of variance (PS-ANOVA) method is used for efficient estimation through mixed model re-parameterization. The results indicate that the phase-division models significantly outperform the benchmark whole-cycle ones either under quadratic utility or exponential utility, given different levels of risk aversions. Finally, regarding hedging effectiveness, the expected utility ratio between the phase-division contract and the whole-cycle contract, and the percentage changes of mean root square loss and variance of revenues support the proposed phase-division contract.