Multi-Armed Bandit

AI and Machine Learning · Reinforcement Learning · also: Bandit problem · multi-armed-bandit.yaml

One state, several arms, unknown payoffs. Strip an MDP of everything except the choice between gathering information and cashing it in, and this is what is left, which is why regret can be bounded here and almost nowhere above it.

• supersedescorrects · extends
• classifiesspecializes · part-of
• substitutes forapproximates · alternative-to
• depends onrequires · validates

Colour is the family; a dashed line is the second member of it.

This node

• specializesis a specific case of Markov Decision Processa single state, so no action changes what is faced next

Referenced by

• part-ofEpsilon-Greedy is a subtopic of this
• part-ofUpper Confidence Bound is a subtopic of this

References

• Reinforcement Learning: An Introduction — Sutton, Barto — 2018 · link
• Background — Multi-armed bandit on Wikipedia