boax.acquisitions.probability_of_improvement

boax.acquisitions.probability_of_improvement#

boax.acquisitions.probability_of_improvement(bounds, improvement_factor=1.0, num_raw_samples=512, num_restarts=10)#

The Probability of Improvement acquisition function.

Probability of improvement over the best function value observed so far.

PI(x) = P(y >= best), y ~ f(x)

Example

>>> acqf = probability_of_improvement(0.2, model)
>>> poi = acqf(xs)
Parameters:

  • best – The best function value observed so far.

  • model – The surrogate model.

Return type:

Acquisition[MultivariateNormal]

Returns:

The corresponding Acquisition.