Thompson Sampling 2D: Belief-Space Bandit Charts

Thompson Sampling is one of the oldest and still best-performing algorithms for the exploration/exploitation trade-off at the heart of reinforcement learning. Instead of a fixed exploration rate like ε-greedy or a hand-tuned confidence bound like UCB1, it keeps a full Bayesian posterior — a Beta(α, β) distribution — over each option's unknown win rate, draws one random sample from every posterior each round, and simply pulls whichever arm's sample is highest. This 2D companion renders those posteriors as overlaid, pannable/zoomable density curves on one shared probability axis and tracks cumulative regret and reward rate on a second linked chart, so you watch the belief distributions narrow and separate purely through chart geometry rather than a rendered 3D scene.