/Q-learning-vs-Sarsa-control-analysis

Simple python scripts used to compare the hyperparameters sensibility in two common reinforcement learning approaches for control (Q-learning & Sarsa) using OpenAI gym's CartPole-v0 environment.

Primary LanguagePythonMIT LicenseMIT

Q-LEARNING vs SARSA control - sensibility analysis

Simple python scripts used to compare the hyperparameters sensibility in two common reinforcement learning approaches for control (Q-learning & Sarsa) using OpenAI gym's CartPole-v0 environment.

Modes

In these scripts you can find a variable called "mode", you can set it to:

  • EPS_SENSIBILITY to analyze the sensibility to the exploration rate;
  • ALPHA_SENSIBILITY to analyze the sensibility to the learning rate;
  • DISCOUNT_SENSIBILITY to analyze the sensibility to the discount rate.

Plots

All plots use the boxplot function from matplotlib library.