algorithms.util

Helper classes that can be used by all learning algorithms.

In particular:

  • action_perturbator can be used to randomly noise (perturb) an action.

  • action_selector can be used to select an action, based on its interest, and a given distribution. These classes target the exploration-exploitation dilemma: finding a trade-off between selecting the action with the best interests, and trying other actions to discover their true interests.

members:

Modules

action_perturbator

This module defines several classes to perturb (explore) actions.

action_selector

This module defines several classes to select actions (ActionSelectors).