Module: tfp.experimental.distributions

TensorFlow Probability experimental distributions package.

Modules

marginal_fns module: Experimental functions to use as marginals for GaussianProcess(es).

Classes

class ImportanceResample: Models the distribution of finitely many importance-reweighted samples.

class IncrementLogProb: A distribution representing an unnormalized measure on a singleton set.

class JointDistributionPinned: A wrapper class for JointDistribution which pins, e.g., the evidence.

class MultiTaskGaussianProcess: Marginal distribution of a Multitask GP at finitely many points.

class MultiTaskGaussianProcessRegressionModel: Posterior predictive in a conjugate Multi-task GP regression model.

class MultivariateNormalPrecisionFactorLinearOperator: A multivariate normal on R^k, parametrized by a precision factor.

Functions

inflated_factory(...): Create Inflated subclasses for specific distributions and positions.

log_prob_ratio(...): Computes p.log_prob(x) - q.log_prob(y), numerically stably.