Elite Fellowship for Data Science Research at KIT

The postdoc, who is a research associate in the group of Prof. Klemens Böhm, will receive up to 150,000 euros over three years from the Baden-Württemberg Stiftung for his project "Sequential Decision Making Algorithms for Knowledge Discovery from Data Streams" to cover personnel, travel, material, and investment costs.

In his project, the scientist explores sequential decision making, or SDM; a fundamental task faced by virtually any agent, i.e., any software system, that interacts with its environment; Simply put, it is the act of answering the question: What should I do next? This task is challenging as the agent must deal with uncertainties coming from the environment and its actions. SDM techniques excel at observing and representing their environment. A common assumption is that, while the underlying environment is uncertain, its nature does not change much, i.e., it is static. This assumption does not hold in general; real-world environments tend to be dynamic.

On the other hand,  it is well-known in the Data Science community that extracting knowledge from data -- the KDD (Knowledge Discovery from Data) process -- is much more complex when data comes as a stream, i.e., data is continuously collected, and its characteristics may change dynamically, in unforeseeable ways. While algorithms for Knowledge Discovery from Data Streams (KDDS) can extract knowledge from dynamic data, existing techniques are somewhat limited in terms of adaptability and applicability.

This research project aims to solve these problems by bridging the gap between KDDS and SDM, as those two topics have mostly been considered separately. Dr. Fouché wants to treat the KDDS process as an SDM process and develop new SDM algorithms for data stream-like environments. This will open new perspectives and give way to unique contributions of general interest in Data Science and numerous real-world applications where data streams are ubiquitous. For example, in materials sciences, one could significantly speed up scientific insights by making optimal data-driven simulation planning, i.e., deciding which simulation(s) are likely to bring the most insights and should be conducted next. Another important application domain is the so-called Industry 4.0, where the interconnection of agents and decentralized decision making plays a vital role.

With its elite program, the Baden-Württemberg Stiftung supports excellent scientists at Baden-Württemberg universities on their way to professorships by enabling them to apply for and manage research projects on their own responsibility.

Particular attention is paid to the networking of postdocs. Several times a year, the Baden-Württemberg Stiftung invites postdocs to attend training and networking meetings focusing on the areas of teaching, science management, or academic self-administration. These meetings allow for discussing important topics, the interdisciplinary exchange of young scientists in similar life situations, and the exchange of experiences with alumni who are already one step further in their scientific careers.

Dr. Edouard Fouché deals with fundamental research questions in the field of Data Science, with a focus on the analysis of data streams. His dissertation, which he defended at KIT in July 2020, was awarded the distinction "Summa Cum Laude'' as well as the Helmholtz Doctoral Award. Dr. Fouché completed two research stays in the USA and Japan as part of his PhD. Before his PhD, he studied computer science in France, his country of origin. Dr. Fouché is a reviewer for worldwide established conferences in his research area (KDD, IJCAI, ICDM, SDM) and recently joined the steering committee of the Karlsruhe House of Young Scientists (KHYS) and the faculty council at the KIT Department of Informatics.