Structure in the Space of Value Functions

  • Authors:
  • David Foster;Peter Dayan

  • Affiliations:
  • Centre for Neuroscience, University of Edinburgh, Edinburgh, UK&semi/ Gatsby Computational Neuroscience Unit, London, UK. djfoster@mit.edu;Gatsby Computational Neuroscience Unit, London. dayan@gatsby.ucl.ac.uk

  • Venue:
  • Machine Learning
  • Year:
  • 2002

Quantified Score

Hi-index 0.00

Visualization

Abstract

Solving in an efficient manner many different optimal control tasks within the same underlying environment requires decomposing the environment into its computationally elemental fragments. We suggest how to find fragmentations using unsupervised, mixture model, learning methods on data derived from optimal value functions for multiple tasks, and show that these fragmentations are in accord with observable structure in the environments. Further, we present evidence that such fragments can be of use in a practical reinforcement learning context, by facilitating online, actor-critic learning of multiple goals MDPs.