tfp.experimental.bayesopt.acquisition.GaussianProcessProbabilityOfImprovement

View source on GitHub

Gaussian Process probability of improvement acquisition function.

Inherits From: AcquisitionFunction

tfp.experimental.bayesopt.acquisition.GaussianProcessProbabilityOfImprovement(
    predictive_distribution, observations, seed=None
)

Computes the analytic sequential probability of improvement for a Gaussian process model relative to observed data.

Requires that predictive_distribution has mean and stddev methods.

Examples

Build and evaluate a GP Probability of Improvement acquisition function.

import numpy as np
import tensorflow_probability as tfp

tfd = tfp.distributions
tfpk = tfp.math.psd_kernels
tfp_acq = tfp.experimental.bayesopt.acquisition

# Sample 10 4-dimensional index points and associated observations.
index_points = np.random.uniform(size=[10, 4])
observations = np.random.uniform(size=[10])

# Build a GP regression model.
dist = tfd.GaussianProcessRegressionModel(
    kernel=tfpk.ExponentiatedQuadratic(),
    observation_index_points=index_points,
    observations=observations)

gp_poi = tfp_acq.GaussianProcessProbabilityOfImprovement(
    predictive_distribution=dist,
    observations=observations)

# Evaluate the acquisition function at a set of predictive index points.
pred_index_points = np.random.uniform(size=[6, 4])
acq_fn_vals = gp_poi(pred_index_points)  # Has shape [6].

Args
predictive_distribution	tfd.Distribution-like, the distribution over observations at a set of index points. Must have mean, stddev methods.
observations	Float Tensor of observations. Shape has the form [b1, ..., bB, e], where e is the number of index points (such that the event shape of predictive_distribution is [e]) and [b1, ..., bB] is broadcastable with the batch shape of predictive_distribution.
seed	PRNG seed; see tfp.random.sanitize_seed for details.

Attributes
is_parallel	Python bool indicating whether the acquisition function is parallel. Parallel (batched) acquisition functions evaluate batches of points rather than single points.
observations	Float Tensor of observations.
predictive_distribution	The distribution over observations at a set of index points.
seed	PRNG seed.

Methods

__call__

View source

__call__(
    **kwargs
)

Computes analytic GP probability of improvement.

Args
**kwargs	Keyword args passed on to the mean and stddev methods of predictive_distribution.

Returns
Probability of improvement at index points implied by predictive_distribution (or overridden in **kwargs).