Samy Bengio

Samy Bengio is a Full Professor at the Ecole Polytechnique Fédérale de Lausanne (EPFL) in the School of Computer and Communication Sciences. He holds multiple roles including Full Professor in the IC Deanship and Lecturer positions in both SSC - Teaching and SIN - Teaching departments, with his office located at INR 140. His academic career spans over three decades with significant contributions to machine learning and artificial intelligence.

Dr. Bengio received his PhD in Computer Science from the University of Montreal in 1993. Following his doctoral studies, he completed three postdoctoral years at CNET (the research center of France Telecom) and INRS-Telecommunications in Montreal. He subsequently worked as a researcher at CIRANO, an academic research center in economics and finance, and served as a research director at Microcell Labs, the research center of a Canadian telecommunications company. Since 1999, he has been a senior researcher and head of the Machine Learning group at IDIAP.

His research interests encompass a broad spectrum of machine learning topics including learning algorithms, automatic speech processing, person authentication, multimodal processing, sequence processing, neural networks, support vector machines, hidden Markov models, and Gaussian mixtures. His work often bridges theoretical foundations with practical applications across various domains.

Analysis of his recent publications (2023-2025) reveals a strong focus on fundamental questions about the reasoning capabilities of neural networks, particularly transformers. His research examines out-of-distribution generalization, the limits of logical reasoning in AI systems, and novel approaches to improve model reasoning through techniques like chain-of-sketch and reversal thinking. A significant portion of his recent work investigates the mathematical foundations of why certain architectures succeed or fail at specific reasoning tasks.

• Outstanding Paper Award at ICML 2023

Dr. Bengio has supervised several doctoral students including Grangier David, Keller Mikaela, Paiement Jean-François, and Poh Norman Hoon Thian. His research group at IDIAP has received substantial funding for projects exploring the theoretical foundations of machine learning and their practical applications in speech recognition, authentication systems, and reasoning models. The group has developed the Torch learning algorithm library, which has been influential in the machine learning community.

As head of the Machine Learning group at IDIAP, Dr. Bengio leads a multidisciplinary team of researchers investigating the fundamental principles of learning algorithms and their applications across various domains including speech processing, computer vision, and natural language understanding. His team collaborates extensively with both academic and industrial partners to advance the state of the art in machine learning.