Domain and Action Changes for Non-Stationary Behavior

[brumfie...@gmail.com]

Hello All,

I am looking into how to train an agent in BURLAP for a number of episodes and then add/remove or change the action parameters. In this case, how is it possible to let a LearningAgent object run learning episodes on with a different domain other than the one declared with the algorithm's class? For example, Q-Learning takes in the domain, the gamma, the hashing factory, the q initialization, and the learning rate.

Or is it as simple as just using the addAction or addState methods in the Episode class?

Marcus B.