Reward is enough for convex MDPs
arXiv:2106.00661
Abstract
Maximising a cumulative reward function that is Markov and stationary, i.e., defined over state-action pairs and independent of time, is sufficient to capture many kinds of goals in a Markov decision process (MDP). However, not all goals can be captured in this manner. In this paper we study convex MDPs in which goals are expressed as convex functions of the stationary distribution and show that they cannot be formulated using stationary reward functions. Convex MDPs generalize the standard reinforcement learning (RL) problem formulation to a larger framework that includes many supervised and unsupervised RL problems, such as apprenticeship learning, constrained MDPs, and so-called `pure exploration'. Our approach is to reformulate the convex MDP problem as a min-max game involving policy and cost (negative reward) `players', using Fenchel duality. We propose a meta-algorithm for solving this problem and show that it unifies many existing algorithms in the literature.
References in corpus (13)
- Variational Intrinsic Control
- Provably Efficient Maximum Entropy Exploration
- A Theory of Regularized Markov Decision Processes
- Online Convex Optimization in Adversarial Markov Decision Processes
- Variational Policy Gradient Method for Reinforcement Learning with General Utilities
- Wasserstein Adversarial Imitation Learning
- Exploration-Exploitation in Constrained MDPs
- Off-Policy Evaluation via the Regularized Lagrangian
- Efficiently Solving MDPs with Stochastic Mirror Descent
- Behavior Priors for Efficient Reinforcement Learning
- Following the Leader and Fast Rates in Linear Prediction: Curved Constraint Sets and Other Regularities
- Concave Utility Reinforcement Learning: the Mean-Field Game Viewpoint
- Towards Tight Bounds on the Sample Complexity of Average-reward MDPs