Instance-Based State Identification for Reinforcement Learning

Abstract

This paper presents instance-based state identification, an approach to reinforcement learning and hidden state that builds disambiguat(cid:173) ing amounts of short-term memory on-line, and also learns with an order of magnitude fewer training steps than several previous ap(cid:173) proaches. Inspired by a key similarity between learning with hidden state and learning in continuous geometrical spaces, this approach uses instance-based (or "memory-based") learning, a method that has worked well in continuous spaces.

Cite

Text

McCallum. "Instance-Based State Identification for Reinforcement Learning." Neural Information Processing Systems, 1994.

Markdown

[McCallum. "Instance-Based State Identification for Reinforcement Learning." Neural Information Processing Systems, 1994.](https://mlanthology.org/neurips/1994/mccallum1994neurips-instancebased/)

BibTeX

@inproceedings{mccallum1994neurips-instancebased,
  title     = {{Instance-Based State Identification for Reinforcement Learning}},
  author    = {McCallum, R. Andrew},
  booktitle = {Neural Information Processing Systems},
  year      = {1994},
  pages     = {377-384},
  url       = {https://mlanthology.org/neurips/1994/mccallum1994neurips-instancebased/}
}