Multiple-Bandit-BatchUpdate

Inspired by the book Bandit Algorithms for Website Optimization.

I extend the code to multiple bandit pproblem (select k arms out of N total arms at each step) as well as batch update process (reward is observed at a batch size of m). Thompson method is also introduced.

Simulation result prefers Thompson Sampling among others.