English

Approximate Temporal Difference Learning is a Gradient Descent for Reversible Policies

Machine Learning 2018-05-03 v1 Optimization and Control Machine Learning

Abstract

In reinforcement learning, temporal difference (TD) is the most direct algorithm to learn the value function of a policy. For large or infinite state spaces, exact representations of the value function are usually not available, and it must be approximated by a function in some parametric family. However, with \emph{nonlinear} parametric approximations (such as neural networks), TD is not guaranteed to converge to a good approximation of the true value function within the family, and is known to diverge even in relatively simple cases. TD lacks an interpretation as a stochastic gradient descent of an error between the true and approximate value functions, which would provide such guarantees. We prove that approximate TD is a gradient descent provided the current policy is \emph{reversible}. This holds even with nonlinear approximations. A policy with transition probabilities P(s,s)P(s,s') between states is reversible if there exists a function μ\mu over states such that P(s,s)P(s,s)=μ(s)μ(s)\frac{P(s,s')}{P(s',s)}=\frac{\mu(s')}{\mu(s)}. In particular, every move can be undone with some probability. This condition is restrictive; it is satisfied, for instance, for a navigation problem in any unoriented graph. In this case, approximate TD is exactly a gradient descent of the \emph{Dirichlet norm}, the norm of the difference of \emph{gradients} between the true and approximate value functions. The Dirichlet norm also controls the bias of approximate policy gradient. These results hold even with no decay factor (γ=1\gamma=1) and do not rely on contractivity of the Bellman operator, thus proving stability of TD even with γ=1\gamma=1 for reversible policies.

Keywords

Cite

@article{arxiv.1805.00869,
  title  = {Approximate Temporal Difference Learning is a Gradient Descent for Reversible Policies},
  author = {Yann Ollivier},
  journal= {arXiv preprint arXiv:1805.00869},
  year   = {2018}
}
R2 v1 2026-06-23T01:42:58.112Z