This repo contains a review of the contextual multi-armed bandits.
Includes proposed framework for extendible building blocks that form the contextual bandit problem.
You can find the overview of the contextual bandits, dataset, and the framework in the presentation
- src/ contains the modules of the framework
- agent, environment, oracle, policy, data provider
- train_manual demonstrates how to combine these modules to form a contextual bandit
- train_grid allows you to run the model for a specified parameter grid