Multi-Armed Bandits & Exploration

The explore/exploit tradeoff, ε-greedy, UCB1, Thompson sampling, and regret — the simplest RL you will actually ship

Learn the multi-armed bandit framework as the simplest form of reinforcement learning used in production: the explore/exploit tradeoff, ε-greedy, UCB1 with a derivation sketch, Thompson sampling with a worked Beta-Bernoulli example, and regret as the metric that makes 'how good is this exploration strategy' precise — contrasted directly with fixed-split A/B testing.

Practice questions (6)

• Choosing Between a Bandit and an A/B Test for Subject Line Selection

  Intermediate · Free
  View →

• Diagnose a Stuck ε-Greedy Policy

  Intermediate
  View →

• Compute and Interpret UCB1 Scores

  Intermediate
  View →

• Run Thompson Sampling by Hand on a Beta-Bernoulli Bandit

  Intermediate
  View →

• Interpret and Compare Regret Curves for Two Deployed Policies

  Advanced
  View →

Quizzes (1)