Machine Learning

, Volume 49, Issue 2, pp 161–178

Kernel-Based Reinforcement Learning

  • Dirk Ormoneit
  • Śaunak Sen
Article

DOI: 10.1023/A:1017928328829

Cite this article as:
Ormoneit, D. & Sen, Ś. Machine Learning (2002) 49: 161. doi:10.1023/A:1017928328829

Abstract

We present a kernel-based approach to reinforcement learning that overcomes the stability problems of temporal-difference learning in continuous state-spaces. First, our algorithm converges to a unique solution of an approximate Bellman's equation regardless of its initialization values. Second, the method is consistent in the sense that the resulting policy converges asymptotically to the optimal policy. Parametric value function estimates such as neural networks do not possess this property. Our kernel-based approach also allows us to show that the limiting distribution of the value function estimate is a Gaussian process. This information is useful in studying the bias-variance tradeoff in reinforcement learning. We find that all reinforcement learning approaches to estimating the value function, parametric or non-parametric, are subject to a bias. This bias is typically larger in reinforcement learning than in a comparable regression problem.

reinforcement learning Markov decision process kernel-based learning kernel smoothing local averaging lazy learning 
Download to read the full article text

Copyright information

© Kluwer Academic Publishers 2002

Authors and Affiliations

  • Dirk Ormoneit
    • 1
  • Śaunak Sen
    • 2
  1. 1.Department of Computer ScienceStanford UniversityStanfordUSA.
  2. 2.The Jackson LaboratoryBar Harbor