Description du projet
Améliorer l’apprentissage par renforcement fondé sur la théorie
L’apprentissage par renforcement (AR) est un sous-domaine de l’apprentissage automatique qui s’intéresse à la façon dont les agents intelligents interagissent avec des environnements inconnus pour maximiser leurs récompenses. L’application potentielle des techniques d’apprentissage par renforcement à des problèmes complexes du monde réel, tels que le contrôle des véhicules autonomes ou les réseaux d’énergie intelligents, a attiré une considérable attention sur ce domaine. Cependant, les algorithmes d’AR les plus avancés ne sont pas applicables dans les domaines les plus prometteurs, en grande partie à cause du manque de garanties formelles de performance. Le projet SCALER, financé par l’UE, entend relever ce défi en adoptant une approche fondée sur des principes pour développer une nouvelle génération d’algorithmes d’apprentissage par renforcement dont l’efficacité et l’évolutivité sont prouvées. La méthodologie sera basée sur l’identification de nouvelles propriétés structurelles des processus de décision de Markov à grande échelle qui permettent un apprentissage efficace sur le plan informatique et statistique.
Objectif
Reinforcement learning (RL) is an intensely studied subfield of machine learning concerned with sequential decision-making problems where a learning agent interacts with an unknown reactive environment while attempting to maximize its rewards. In recent years, RL methods have gained significant popularity due to being the key technique behind some spectacular breakthroughs of artificial intelligence (AI) research, which renewed interest in applying such techniques to challenging real-world problems like control of autonomous vehicles or smart energy grids. While the RL framework is clearly suitable to address such problems, the applicability of the current generation of RL algorithms is limited by a lack of formal performance guarantees and a very low sample efficiency. This project proposes to address this problem and advance the state of the art in RL by developing a new generation of provably efficient and scalable algorithms. Our approach is based on identifying various structural assumptions for Markov decision processes (MDPs, the main modeling tool used in RL) that enable computationally and statistically efficient learning. Specifically, we will focus on MDP structures induced by various approximation schemes including value-function approximation and relaxations of the linear-program formulation of optimal control in MDPs. Based on this view, we aim to develop a variety of new tools for designing and analyzing RL algorithms, and achieve a deep understanding of fundamental performance limits in structured MDPs. While our main focus will be on rigorous theoretical analysis of algorithms, most of our objectives are inspired by practical concerns, particularly by the question of scalability. As a result, we expect that our proposed research will have significant impact on both the theory and practice of reinforcement learning, bringing RL methods significantly closer to practical applicability.
Champ scientifique
- engineering and technologymechanical engineeringvehicle engineeringautomotive engineeringautonomous vehicles
- engineering and technologyenvironmental engineeringenergy and fuelsrenewable energy
- natural sciencescomputer and information sciencesartificial intelligencemachine learningreinforcement learning
Programme(s)
Thème(s)
Régime de financement
ERC-STG - Starting GrantInstitution d’accueil
08002 Barcelona
Espagne