Reinforcement learning and Knowledge-driven AI: Within the junior research group “Nature-Inspired Machine Learning”, the team is investigating the connection of knowledge graphs and large language models, and the role of Neuro-symbolic methods for future learning models. This is tighten with reinforcement learning approaches which is also a major interest for the group.

Learning and Reasoning: Another focal research direction of the group “Nature-Inspired Machine Learning” are link prediction techniques, escpecially to provide recommendations in the context of scholarly communication. Furthermore, the usage of knowledge graph embedding (KGE) models in several other use cases, as well proposing new models, is an important part of the research.

Especially, this junior research group is focusing on following key topics:

• Machine Learning
• Foundation Models
• Representation Learning
• Reinforcement Learning

Driven by a deep passion for advancing the frontiers of knowledge, the overarching goal is to
cultivate an exceptional research ecosystem that empowers emerging scholars to conduct
groundbreaking and innovative research. Vahdati and her group are committed to devising strategies that not only enhance the wellbeing of individuals—encompassing both their physical and mental
health—but also contribute positively to environmental sustainability through the ethical
application of artificial intelligence. The groups’s vision is centered on creating tangible, positive
changes in society by harnessing the potential of cutting-edge research to improve lives and
protect our planet.

In fact, the research addresses critical global challenges by improving human health, advancing
environmental sustainability, and applying AI ethically to foster positive societal change. It’s
relevant because it offers practical solutions to enhance well-being and protect our planet,
aligning with urgent contemporary needs and future priorities.

Projects

By now, there are no projects dedicated to ScaDS.AI explicitly but are part of Vahdati’s research at the InfAI. Until the end of 2027, the EU project “IntelliLung” focuses on artificial intelligence to help optimize ventilation for intensive care patients. Vahdati has been directing the group research activities, and
research mainly on her vision for research in machine intelligence inspired by natural science
that can result in innovations to address weaknesses of current ML approaches. The main
activities of the group and planned research directions are focused on representation
learning.

More information can be found on the official website of the project.

Team

Lead

The junior research group leader, Dr. Sahar Vahdati, started at ScaDS.AI Dresden on 01.09.2023. Vahdati is a PhD graduate from University of Bonn, on Intelligent Information Systems track, particularly on Representation Learning Over Graphs. At the Institute for Computer Science III, she die her Ph.D. studies in the Smart Data Analytics Group. Vahdati wrote her Ph.D. thesis “Collaborative Integration, Publishing, and Analysis of Distributed Scholarly Metadata” in the research area a of Semantic Web and Linked Data, Knowledge Graph Integration and Mining.

Previously, Vahdati was leading the NIMI (“Nature-Inspired Machine Intelligence”) research group at the Institute for Applied Informatics (InfAI). The group was built from scratch within the Efficient Technology Integration competence center at InfAI. Also, in her previous position as a Postdoc at Oxford University in the Information Systems group led by Prof. Georg Gottlob, Vahdati explored logic-based learning and reasoning over graphs.

Vahdati’s scientific achievements are very diverse: By now, she acquired 1.54 mio Euro third party funding over 3 years. She has co-authored 90+ publications, which appeared, for example, at top AI conferences such as AAAI, PAKDD, IJCNN, EMNLP. The articles were cited over 800 times, according to Google Scholar, at the time of writing this profile details.
Teaching and mentoring researchers at different levels of their career is a very central activity for Vahdati. In Oxford as well as in Bonn, she was involved in teaching Datalog and Logic-based reasoning, Semantic Web and Learning over Knowledge Graphs. Currently, Vahdati is co-mentoring eight PhD candidates across several organizations, as well as two postdoctoral fellows in her group, and several master students.

More information about Dr. Sahar Vahdati and her work can be found on her personal website.

Team Members

The team members of this group are both doing research on large language models.

Publications

By now, no papers by the junior research group exist.