Thompson Sampling — Bayesian bandit that learns from your preferences using probability distributions