Sparse Distributed Memories for On-Line Value-Based Reinforcement Learning
Abstract
In this paper, we advocate the use of Sparse Distributed Memories (SDMs) for on-line, value-based reinforcement learning (RL). SDMs provide a linear, local function approximation scheme, designed to work when a very large/ high-dimensional input (address) space has to be mapped into a much smaller physical memory. We present an implementation of the SDM architecture for on-line, value-based RL in continuous state spaces. An important contribution of this paper is an algorithm for dynamic on-line allocation and adjustment of memory resources for SDMs, which eliminates the need for choosing the memory size and structure a priori. In our experiments, this algorithm provides very good performance while efficiently managing the memory resources.
Cite
Text
Ratitch and Precup. "Sparse Distributed Memories for On-Line Value-Based Reinforcement Learning." European Conference on Machine Learning, 2004. doi:10.1007/978-3-540-30115-8_33Markdown
[Ratitch and Precup. "Sparse Distributed Memories for On-Line Value-Based Reinforcement Learning." European Conference on Machine Learning, 2004.](https://mlanthology.org/ecmlpkdd/2004/ratitch2004ecml-sparse/) doi:10.1007/978-3-540-30115-8_33BibTeX
@inproceedings{ratitch2004ecml-sparse,
title = {{Sparse Distributed Memories for On-Line Value-Based Reinforcement Learning}},
author = {Ratitch, Bohdana and Precup, Doina},
booktitle = {European Conference on Machine Learning},
year = {2004},
pages = {347-358},
doi = {10.1007/978-3-540-30115-8_33},
url = {https://mlanthology.org/ecmlpkdd/2004/ratitch2004ecml-sparse/}
}