Policy Gradient Methods for Reinforcement Learning with Function Approximation
Richard S. Sutton, David McAllester, Satinder Pal Singh et al. · 1999
Function approximation is essential to reinforcement learning, but the standard approach of approximating a value function and determining a policy from it has so far proven theoretically intractable. In this paper we explore an alternative approach in which the policy is explicitly represented by its own function approximator, independent of the value function, and is updated according to the gradient of expected reward with respect to the policy parameters. Williams's REINFORCE method and actor--critic methods are examples of this approach. Our main new result is to show that the gradient can be written in a form suitable for estimation from experience aided by an approximate action-value or advantage function. Using this result, we prove for the first time that a version of policy iteration with arbitrary di#erentiable function approximation is convergent to a locally optimal policy. Large applications of reinforcement learning (RL) require the use of generalizing function approxima...
Cited by 4,562 works.
Authors
References
- Reinforcement Learning: An Introduction
- Neuronlike adaptive elements that can solve difficult learning control problems
- Introduction to Reinforcement Learning
- Simple statistical gradient-following algorithms for connectionist reinforcement learning
- Neuro-dynamic Programming
- Temporal credit assignment in reinforcement learning
- Residual Algorithms: Reinforcement Learning with Function Approximation
- Actor-critic algorithms
- Stable Function Approximation in Dynamic Programming
- Feature-based methods for large scale dynamic programming