Upper Confidence Bound (UCB)

Upper Confidence Bound (UCB) is an acquisition function that scores candidates based on both their predicted mean and uncertainty. It selects candidates with high upper confidence bounds, balancing exploitation and exploration through a tunable parameter.

class alf_tools.optimizer.acquisition_functions.ucb.UCB(alpha)

Bases: AcquisitionFunction

Upper Confidence Bound acquisition function.

The UCB acquisition value is given by: UCB = μ + ασ, where μ is the mean prediction, σ is the standard deviation, and α is the exploration parameter. Higher α values lead to more exploration, while lower α values lead to more exploitation. This is a maximising acquisition function.