Multi-Armed Bandits

  1. The Casino

    Random, epsilon-greedy and UCB strategies compete over a row of slot machines.

  2. Regret

    Measuring the cost of bad decisions as opportunity loss.

  3. Dynamic Environments

    What happens to each strategy when the world changes underneath it.

  4. Contextual Bandits

    Segmenting an audience and running an experiment per context.

  5. The Tyranny of Averages

    Why iterative A/B testing is the wrong approach to user experience optimization.

  6. Variant Iteration: Finding New Best Variants

    Coming later.

  7. Sequential Decision Making

    Coming later.

Reinforcement Learning