reinforcelearn: Reinforcement Learning

Implements reinforcement learning environments and algorithms as described in Sutton & Barto (1998, ISBN:0262193981). The Q-Learning algorithm can be used with function approximation, eligibility traces (Singh & Sutton (1996) <doi:10.1007/BF00114726>) and experience replay (Mnih et al. (2013) <arXiv:1312.5602>).

Version: 0.2.1
Depends: R (≥ 3.0.0)
Imports: checkmate (≥ 1.8.4), R6 (≥ 2.2.2), nnet (≥ 7.3-12), purrr (≥ 0.2.4)
Suggests: reticulate, keras, knitr, rmarkdown, testthat, covr, lintr
Published: 2019-04-09
Author: Markus Dumke [aut, cre]
Maintainer: Markus Dumke <markusdumke at>
License: MIT + file LICENSE
NeedsCompilation: no
SystemRequirements: (Python and gym only required if gym environments are used)
Materials: README NEWS
CRAN checks: reinforcelearn results


Reference manual: reinforcelearn.pdf
Vignettes: Agents
Package source: reinforcelearn_0.2.1.tar.gz
Windows binaries: r-devel:, r-release:, r-oldrel:
macOS binaries: r-release: reinforcelearn_0.2.1.tgz, r-oldrel: reinforcelearn_0.2.1.tgz
Old sources: reinforcelearn archive


Please use the canonical form to link to this page.