distribution_math

bojaxns.gaussian_process_formulation.distribution_math

Module Contents

tfpd[source]
log_normal(x, mean, cov)[source]
log_normal_with_mask(x, mean, cov, sigma)[source]

Computes log-Normal density in a numerically stable way so that sigma can contain +inf for masked data.

Parameters:

  • x – RV value

  • mean – mean of Gaussian

  • cov – covariance of underlying, minus the obs. covariance

  • sigma – stddev’s of obs. error, inf encodes an outlier.

Returns: a normal density for all points not of inf stddev obs. error.

class GaussianProcessData[source]

Bases: NamedTuple

U: jax.numpy.ndarray[source]
Y: jax.numpy.ndarray[source]
Y_var: jax.numpy.ndarray[source]
sample_size: jax.numpy.ndarray[source]
exception NotEnoughData[source]

Bases: Exception

Common base class for all non-exit exceptions.

Initialize self. See help(type(self)) for accurate signature.

class GaussianProcessConditionalPredictive(data, kernel, variance, mean)[source]

Bases: bojaxns.base.ConditionalPredictive

Parameters:

  • data (GaussianProcessData) –

  • kernel (tensorflow_probability.substrates.jax.math.psd_kernels.PositiveSemidefiniteKernel) –

  • variance (jax.numpy.ndarray) –

  • mean (jax.numpy.ndarray) –

posterior()[source]
marginal_likelihood()[source]
__call__(U_star, cov=False)[source]
Parameters:

  • U_star (jax.numpy.ndarray) –

  • cov (bool) –

class GaussianProcessConditionalPredictiveFactory(data)[source]

Bases: bojaxns.base.ConditionalPredictiveFactory

Parameters:

data (GaussianProcessData) –

ndims()[source]
build_prior_model()[source]
psd_kernels()[source]
Return type:

List[Type[tensorflow_probability.substrates.jax.math.psd_kernels.PositiveSemidefiniteKernel]]

__call__(**samples)[source]
Return type:

GaussianProcessConditionalPredictive

class ExpectedImprovementAcquisition(conditional_predictive)[source]

Bases: bojaxns.base.AbstractAcquisition

A class that represents the heteroscedastic expected improvement acquisition function.

Parameters:

conditional_predictive (GaussianProcessConditionalPredictive) –

__call__(u_star)[source]
Parameters:

u_star (jax.numpy.ndarray) –

class ScaledExpectedImprovementAcquisition(condition_predictive)[source]

Bases: bojaxns.base.AbstractAcquisition

A class that represents the heteroscedastic expected improvement acquisition function.

Parameters:

condition_predictive (GaussianProcessConditionalPredictive) –

__call__(u_star)[source]
Parameters:

u_star (jax.numpy.ndarray) –

class TopTwoAcquisition(condition_predictive, u1)[source]

Bases: bojaxns.base.AbstractAcquisition

A class that represents any acquisition function. All acquisition functions take a point in the U-domain and returns a metric that gives a proxy as to how valuable it would be to try that point. All acquisition values only make sense relatively.

Parameters:
__call__(u_star)[source]
Parameters:

u_star (jax.numpy.ndarray) –

class ScaledTopTwoAcquisition(condition_predictive, u1)[source]

Bases: bojaxns.base.AbstractAcquisition

A class that represents any acquisition function. All acquisition functions take a point in the U-domain and returns a metric that gives a proxy as to how valuable it would be to try that point. All acquisition values only make sense relatively.

Parameters:
__call__(u_star)[source]
Parameters:

u_star (jax.numpy.ndarray) –

class ExpectedImprovementAcquisitionFactory(conditional_predictive_factory)[source]

Bases: bojaxns.base.AcquisitionFactory

Parameters:

conditional_predictive_factory (GaussianProcessConditionalPredictiveFactory) –

__call__(**sample)[source]
Return type:

bojaxns.base.AbstractAcquisition

class ScaledExpectedImprovementAcquisitionFactory(conditional_predictive_factory)[source]

Bases: bojaxns.base.AcquisitionFactory

Parameters:

conditional_predictive_factory (GaussianProcessConditionalPredictiveFactory) –

__call__(**sample)[source]
Return type:

bojaxns.base.AbstractAcquisition

class TopTwoAcquisitionFactory(conditional_predictive_factory, u1)[source]

Bases: bojaxns.base.AcquisitionFactory

Parameters:
__call__(**sample)[source]
Return type:

bojaxns.base.AbstractAcquisition

class ScaledTopTwoAcquisitionFactory(conditional_predictive_factory, u1)[source]

Bases: bojaxns.base.AcquisitionFactory

Parameters:
__call__(**sample)[source]
Return type:

bojaxns.base.AbstractAcquisition