Abstract | ||
---|---|---|
Many recent successful (deep) reinforcement learning algorithms make use of regularization, generally based on entropy or on Kullback-Leibler divergence. We propose a general theory of regularized Markov Decision Processes that generalizes these approaches in two directions: we consider a larger class of regularizers, and we consider the general modified policy iteration approach, encompassing both policy iteration and value iteration. The core building blocks of this theory are a notion of regularized Bellman operator and the Legendre-Fenchel transform, a classical tool of convex optimization. This approach allows for error propagation analyses of general algorithmic schemes of which (possibly variants of) classical algorithms such as Trust Region Policy Optimization, Soft Q-learning, Stochastic Actor Critic or Dynamic Policy Programming are special cases. This also draws connections to proximal convex optimization, especially to Mirror Descent. |
Year | Venue | Field |
---|---|---|
2019 | arXiv: Learning | Trust region,Mathematical optimization,Divergence,Propagation of uncertainty,Computer science,Markov decision process,Regularization (mathematics),Operator (computer programming),Artificial intelligence,Convex optimization,Machine learning,Reinforcement learning |
DocType | Volume | Citations |
Journal | abs/1901.11275 | 0 |
PageRank | References | Authors |
0.34 | 25 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Matthieu Geist | 1 | 385 | 44.31 |
Bruno Scherrer | 2 | 126 | 14.58 |
Olivier Pietquin | 3 | 664 | 68.60 |