pycalib.gp_classes Module

Gaussian Process classes enabling variational inference of a single GP in GPcalibration.

Classes

DataHolder(value[, dtype, fix_shape, name])	DataHolder is similar to the Parameter with only difference that default values for fix_shape and trainable options are opposite to the Parameter’s and it does not have prior and transform options.
GPModel(X, Y, kern, likelihood, mean_function)	A base class for Gaussian process models, that is, those of the form
Log()	Natural logarithm prior mean function.
MeanFunction([name])	The base mean function class.
Minibatch(value[, batch_size, shuffle, …])	Minibatch is a special case of data holders.
MultiCal(num_classes[, invlink, …])
Parameter(value[, transform, prior, …])	Parameter class is a cornerstone of the GPflow package.
Parameterized([name])	Parameterized object represents a set of computations over children nodes and one of the main purposes is to store these children node like objects.
SVGPcal(X, Y, kern, likelihood[, feat, …])	Probability calibration using a sparse variational latent Gaussian process.
ScalarMult([alpha])	Scalar multiplication mean function.
SoftArgMax(num_classes, **kwargs)	This class implements the multi-class softargmax inverse-link function.