Solving Multi-Armed Problem

Repository consists of:

• Notebook (mab_problem/notebook/multi_armed_bandit.ipynb) with explanations on how to deal with multi-armed problems through four different approaches:

1. Random Selection
2. Epsilon Greedy
3. Thompson Sampling
4. Upper Confidence Bound (UCB1)

Should be opened by Jupyter NBViewer in order to see the plots.

• Flask app (mab_problem/flask_app) for interactive experience with 2 variants and 1000 trials.

App preview:

To run an app on your machine clone/download the repo and follow the commands:

$ cd mab_problem/flask_app
$ export FLASK_APP=app.py
$ flask run