Finite Sample Analyses for TD(0) with Function Approximation

Abstract

TD(0) is one of the most commonly used algorithms in reinforcement learning. Despite this, there is no existing finite sample analysis for TD(0) with function approximation, even for the linear case. Our work is the first to provide such results. Existing convergence rates for Temporal Difference (TD) methods apply only to somewhat modified versions, e.g., projected variants or ones where stepsizes depend on unknown problem parameters. Our analyses obviate these artificial alterations by exploiting strong properties of TD(0). We provide convergence rates both in expectation and with high-probability. The two are obtained via different approaches that use relatively unknown, recently developed stochastic approximation techniques.

Cite

Text

Dalal et al. "Finite Sample Analyses for TD(0) with Function Approximation." AAAI Conference on Artificial Intelligence, 2018. doi:10.1609/AAAI.V32I1.12079

Markdown

[Dalal et al. "Finite Sample Analyses for TD(0) with Function Approximation." AAAI Conference on Artificial Intelligence, 2018.](https://mlanthology.org/aaai/2018/dalal2018aaai-finite/) doi:10.1609/AAAI.V32I1.12079

BibTeX

@inproceedings{dalal2018aaai-finite,
  title     = {{Finite Sample Analyses for TD(0) with Function Approximation}},
  author    = {Dalal, Gal and Szörényi, Balázs and Thoppe, Gugan and Mannor, Shie},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year      = {2018},
  pages     = {6144-6160},
  doi       = {10.1609/AAAI.V32I1.12079},
  url       = {https://mlanthology.org/aaai/2018/dalal2018aaai-finite/}
}