Thompson Sampling Multi-Armed Bandit

NumPy implementation of Thompson Sampling for K-actions stochastic bandit with a normal reward distribution
Further Optimal Regret Bounds for Thompson Sampling by Shipra Agrawal and Navin Goyal is also implemented in a specific part of the code https://arxiv.org/abs/1209.3353
required python libraries: numpy ,matplotlib ,random, stats
.ipynb file with ploted results has been added to the repo

part 1- 4 arms bandit for horizon 100 using Thompson sampling:

part 2- 2 arms bandit for horizon 100 using Thompson sampling σ2 = 0.25 and μ is uniformly sampled in the interval [0, 10]:

part 3- TS-Normal strategy by Agrawal & Goyal - 2 arm bandit m =0

part 3- TS-Normal strategy by Agrawal & Goyal - 2 arm bandit m =5

part 3- TS-Normal strategy by Agrawal & Goyal - 2 arm bandit m =10