boax.acquisitions module

Implements functionalities to construct acquisition functions.

boax.acquisitions

Acquisition Types

class boax.acquisitions.Acquisition(*args, **kwargs)

A callable type for acquisition functions.

An acquisition function takes a set of index_points as input and returns a numeric acquisition value.

Acquisitions

Analytic Acquisitions

probability_of_improvement(bounds[, ...])	The Probability of Improvement acquisition function.
log_probability_of_improvement(bounds[, ...])	The Log Probability of Improvement acquisition function.
expected_improvement(bounds[, ...])	The Expected Improvement acquisition function.
log_expected_improvement(bounds[, ...])	The Log Expected Improvement acquisition function.
upper_confidence_bound(bounds[, beta, ...])	The Upper Confidence Bound (UCB) acquisition function.

Monte Carlo Acquisitions

q_probability_of_improvement(bounds, ...[, ...])	MC-based batch Probability of Improvement acquisition function.
q_expected_improvement(bounds, batch_size, ...)	MC-based batch Expected Improvement acquisition function.
q_upper_confidence_bound(bounds, batch_size, ...)	MC-based batch Upper Confidence Bound acquisition function.

boax.acquisitions.surrogates

Surrogate Types

class boax.acquisitions.surrogates.Surrogate(init, update, prior, posterior, best)

Surrogates

single_task_gaussian_process(bounds[, ...])

The single task gaussian process surrogate model.