[DMANET] [Habilitation Thesis Defense] Martin Krejca

Dear all, It is my great pleasure to invite you to my habilitation thesis defense on *October 9 (Friday), 2026* from *12 p.m. (noon, Paris time) onward* in *room 26-00/534* of the *Jussieu campus*, Paris, France. The thesis is titled /Understanding Complex Behavior Emerging From Randomized Processes in Artificial Intelligence/ and is online available at the following link. The abstract and the jury are at the end of this e-mail. https://www.lix.polytechnique.fr/Labo/Martin.KREJCA/files/hdr-thesis/emergence-of-complexity-in-ai.pdf You can *join online* with the following link: *https://rendez-vous.renater.fr/private/HDR_Defense_Krejca_id_53539c99cefc1de44b49accbacad2bb3c72d62f8218aaccd66bf8908caae71cc__uws-7r9fkgbr_b9d93e-f5f9e7-8ca614* *If you intend to join online*, please contact me (martin.krejca@polytechnique.edu) so that I can provide you with the password that is required for joining. *If you intend to join on-site*, then please register with the following link by October 4. Space is /immensely/ scarce, limited to 8 seats, so I will need to make a selection if the demand is too high. If possible, please prefer to join online. https://evento.renater.fr/survey/martin-krejca-hdr-defense-ih885iad If you wish the join the celebration after the defense, then please register with the following link by October 4. https://evento.renater.fr/survey/martin-krejca-hdr-celebration-hlkoawhv I am happy to see you both online or on-site! Best regards, Martin *Abstract* Artificial intelligence (AI) is a vast area in computer science that spans various research directions. Many of these directions analyze or improve state-of-the-art models, forming the /core/ of modern AI research. The remaining directions investigate less prominent AI approaches, methods supporting the state of the art, or topics that are also studied using modern AI models. These constitute the /wider/ AI area and provide an important extension of the core. In this thesis, I summarize my scientific contributions to two fields that fall into the wider AI area: multi-objective evolutionary algorithms (MOEAs) and stochastic infection processes. MOEAs are optimization heuristics that construct complex solutions to hard problems by iteratively applying simple operations—a concept also present in other AI methods. Moreover, evolutionary principles are used in order to enhance existing AI approaches, acting as supporting tools. On the other hand, stochastic infection processes model spreading phenomena in networks, which are frequently subject to investigation via modern AI tools. The processes underlying both of these areas are similar, with simple iterative decisions resulting in overall complex behavior. This allows studying them via similar approaches, which forms the basis of my research. My main goal is to derive mathematically rigorous guarantees for such processes that provide deep insights for the reasons of the complex behavior that we observe. Since my PhD, my line of research in these two areas has led to several publications, most of them at top AI venues. In this thesis, I discuss these contributions, describe how I use these topics to educate a new generation of researchers, and I provide an overview of my vision for how I will contribute to these areas in the future. For MOEAs, my contributions analyze the impact of different operations on the performance of state-of-the-art algorithms. We find that some operations have issues or bottlenecks, and we provide better alternatives that can provably result in drastic performance improvements. For stochastic infection processes, we pioneer the fully rigorous analysis of an established model where infected agents can be temporarily immune, and we study the influence of diminishing and of increasing infection rates. Notably, we find that the exact definition of immunity plays a crucial role in the resulting behavior. *Jury* *Christoph Dürr*, Sorbonne University, France, /Head of jury/ *Joshua Knowles*, Schlumberger Cambridge Research, UK, /Reviewer/ *Johannes Lengler*, ETH Zurich, Switzerland, /Reviewer/ *Jonathan E. Rowe,* University of Birmingham, UK, /Reviewer/ *Carlos A. Coello Coello*, CINVESTAV-IPN, Mexico, /Jury member/ *Marc Schoenauer*, INRIA, France, /Jury member/ ********************************************************** * * Contributions to be spread via DMANET are submitted to * * DMANET@zpr.uni-koeln.de * * Replies to a message carried on DMANET should NOT be * addressed to DMANET but to the original sender. The * original sender, however, is invited to prepare an * update of the replies received and to communicate it * via DMANET. * * DISCRETE MATHEMATICS AND ALGORITHMS NETWORK (DMANET) * http://www.zaik.uni-koeln.de/AFS/publications/dmanet/ * **********************************************************