TaxiTrainer is a tool to try Reinforcement Learning easily on a taxi problem. The driver must pick up the passenger at the blue letter and drop him at the purple letter.
There is an example of code at the bottom of the script, and the different parameters you can play with are :
- Alpha : learning rate
- Gamma : discount_rate
- Epsilon : The probability that a random action will be picked over the best action according to the model's policy.
- n_epochs : The number of games that will be simulated for training
- render : Set it to true to see the training happening.