Fast Feature Selection for Linear Value Function Approximation

Authors

  • Bahram Behzadian University of New Hampshire
  • Soheil Gharatappeh University of New Hampshire
  • Marek Petrik University of New Hampshire

Abstract

Linear value function approximation is a standard approach to solving reinforcement learning problems with large state spaces. Since designing good approximation features is difficult, automatic feature selection is an important research topic. We propose a new method for feature selection that is based on a low-rank factorization of the transition matrix. Our approach derives features directly from high-dimensional raw inputs, such as image data. The method is easy to implement using SVD, and our experiments show that it is faster and more stable than alternative methods.

Downloads

Published

2021-05-25

How to Cite

Behzadian, B., Gharatappeh, S., & Petrik, M. (2021). Fast Feature Selection for Linear Value Function Approximation. Proceedings of the International Conference on Automated Planning and Scheduling, 29(1), 601-609. Retrieved from https://ojs.aaai.org/index.php/ICAPS/article/view/3527