Method
Multi-armed bandit
Adaptive allocation with Thompson sampling, epsilon-greedy and contextual bandits.
40 cases from 25 companies apply Multi-armed bandit. 18 were checked in an online A/B test; 5 report results in production. Includes variants: Thompson sampling.
Typical tasks: Recommend to new users and new items, Optimize decisions, Avoid false wins in A/B tests, Rank search results and 12 more. Companies: DoorDash, Target, Netflix, Instacart, Lyft, Ramp and 19 more. Often combined with: Thompson sampling, A/B test, Gradient boosting, Rule-based heuristics, Doubly robust estimation and 66 more.



