momtr / Q-Learning

🚀 Q Learning algorithm

QLearning

Implementation of the Q-Learning algorithm.

Todo

• QL - algorithm
• DeepQ Learning
• more examples
• SARSA algorithm

Run

Start a server an run index.html.

Example One

Think of following grid world / environment:

Here, R is the Agent in the environment. Its goal is to reach the terminal state G by moving from one field to another. The agent must not go into the traps denoted by '---'.

Actions

• 0: NORTH
• 1: SOUTH
• 2: WEST
• 3: EAST

Rewards

• +1000 at G
• -1000 at trap (---)
• -1 any other field
• -10 if he moves out of the world

Result

The agent is able to solve the problem with the minimum number of moves.