Algorithms to Live By: The Computer Science of Human Decisions

by Brian Christian

In a multi-armed bandit problem, an Upper Confidence Bound algorithm says, quite simply, to pick the option for which the top of the confidence interval is highest.