Q-learning hyperparameters – alpha, gamma, and epsilon

The hyperparameters of your model are the parameters that are external to the model that you will set yourself (think of the  value in an algorithm such as -nearest neighbors; this is a hyperparameter. You have to set it yourself as there is nothing inherent in the data that determines what  should be).

The three most important hyperparameters for your agent are as follows:

• Alpha: The learning rate
• Gamma: The discount rate
• Epsilon: The exploration rate