Bienvenue sur la page du groupe de travail en apprentissage profond du Laboratoire de Mathématiques de Lens.
On se réunit le mardi à 13h30 en salle P-108.
21/10/25 - Nicola Carissimi (Université de Lille) : Le cube de Rubik via LLM et GRPO
Origine du GRPO : arXiv:2402.03300
14/10/25 - Alexander Chervov (Intitut Curie) : "CayleyPy - Artificial intelligence methods for group and graph theories"
Abstract: We will present an AI-based open source Python library "CayleyPy" which can handle googol-size Cayley graphs and significantly outperforms classical computer algebra systems GAP/SAGE for several tasks. Hundreds conjectures and several results were obtained with its help. Classical group theory tasks e.g. decomposition of the group elements, can be rephrased as standard reinforcement learning tasks, and approached in a similar manner as Google Deepmind's AlphaGo/Zero. We will also give an overview of various recent achievements in "AI for math" emerging field of research.
The talk will be based on: arXiv:2509.19162, arXiv:2502.18663, arXiv:2502.13266.
07/10/25 - Ryuichiro Hataya (Kyoto University) : "Investigating Transformers' Optimization and Adaptation Mechanisms"
Abstract: Transformers are the essential backbone of today's AI systems. This talk investigates the theoretical mechanisms behind their success in training and adaptation. It will cover their optimization dynamics, domain adaptation, and in-context learning capabilities.
30/09/25 - Ivo Dell'Ambrogio (LML) : Nano GPT 4 // Germain Poloudny (LML) : sujet de thèse
23/09/25 - Ivo Dell'Ambrogio (LML) : Nano GPT 3
16/09/25 - Ivo Dell'Ambrogio (LML) : Nano GPT 2
09/09/25 - Ivo Dell'Ambrogio (LML) : Nano GPT 1
Vision et discussion de la vidéo de Andrej Karpathy "":
https://www.youtube.com/watch?v=kCc8FmEb1nY
Ressources :