The Values of Actions in Reinforcement Learning using Q-learning

The Q-learning algorithm implemented from scratch in Python

Agent traversing a maze; GIF by author

This article is a continuation of a series of articles about Reinforcement Learning (RL). Check out the other articles here:

First Steps in the World Of Reinforcement Learning using Python

Temporal Differences with Python — First Sample-Based Reinforcement Learning Algorithm

All the codes used can be viewed here: https://github.com/Eligijus112/rl-snake-game

The notebook with all the plotting functions and agent training codes can be viewed here: https://github.com/Eligijus112/rl-snake-game/blob/master/chapter-6-qlearning.ipynb

In this article, I will present the reader with the concept of Q-values. For the sake of intuition, the reader can change the Q in Q-values for Quality-values. The q values are numeric values that assign a score for each action taken from each state: