Gym Reinforce

Experiment panel for trying reinforcement learning algorithms with some tricks.

Prerequisites

You need Miniconda or typical Anaconda to setup an environment.

  $ conda env create -f environment.yml
  $ conda activate gym

A sample car racing game, spun off from https://github.com/frankfurt-hackathon/carracing-aigym is implemented with Temporal-Difference learning. This can be run as follows.

  $ python3 -m game.carrace

Gym will start up a simulation environment and game module will govern the learning loop. See model/ for the output agent and its learning log.

BSD-2