View source on GitHub
|
Probabilistic Layers.
Modules
conv_variational module: Convolutional variational layers.
dense_variational module: Dense variational layers.
dense_variational_v2 module: DenseVariational layer.
distribution_layer module: Layers for combining tfp.distributions and tf.keras.
initializers module: Keras initializers useful for TFP Keras layers.
masked_autoregressive module: Layers for normalizing flows and masked autoregressive density estimation.
util module: Utilities for probabilistic layers.
variable_input module: VariableInputLayer.
weight_norm module: Layer wrapper for weight normalization.
Classes
class AutoregressiveTransform: An autoregressive normalizing flow layer.
class BlockwiseInitializer: Initializer which concats other intializers.
class CategoricalMixtureOfOneHotCategorical: A OneHotCategorical mixture Keras layer from k * (1 + d) params.
class Convolution1DFlipout: 1D convolution layer (e.g. temporal convolution) with Flipout.
class Convolution1DReparameterization: 1D convolution layer (e.g. temporal convolution).
class Convolution2DFlipout: 2D convolution layer (e.g. spatial convolution over images) with Flipout.
class Convolution2DReparameterization: 2D convolution layer (e.g. spatial convolution over images).
class Convolution3DFlipout: 3D convolution layer (e.g. spatial convolution over volumes) with Flipout.
class Convolution3DReparameterization: 3D convolution layer (e.g. spatial convolution over volumes).
class DenseFlipout: Densely-connected layer class with Flipout estimator.
class DenseLocalReparameterization: Densely-connected layer class with local reparameterization estimator.
class DenseReparameterization: Densely-connected layer class with reparameterization estimator.
class DenseVariational: Dense layer with random kernel and bias.
class DistributionLambda: Keras layer enabling plumbing TFP distributions through Keras models.
class IndependentBernoulli: An Independent-Bernoulli Keras layer from prod(event_shape) params.
class IndependentLogistic: An independent logistic Keras layer.
class IndependentNormal: An independent normal Keras layer.
class IndependentPoisson: An independent Poisson Keras layer.
class KLDivergenceAddLoss: Pass-through layer that adds a KL divergence penalty to the model loss.
class KLDivergenceRegularizer: Regularizer that adds a KL divergence penalty to the model loss.
class MixtureLogistic: A mixture distribution Keras layer, with independent logistic components.
class MixtureNormal: A mixture distribution Keras layer, with independent normal components.
class MixtureSameFamily: A mixture (same-family) Keras layer.
class MultivariateNormalTriL: A d-variate MVNTriL Keras layer from d + d * (d + 1) // 2 params.
class OneHotCategorical: A d-variate OneHotCategorical Keras layer from d params.
class VariableLayer: Simply returns a (trainable) variable, regardless of input.
class VariationalGaussianProcess: A VariationalGaussianProcess Layer.
Functions
default_loc_scale_fn(...): Makes closure which creates loc, scale params from tf.get_variable.
default_mean_field_normal_fn(...): Creates a function to build Normal distributions with trainable params.
default_multivariate_normal_fn(...): Creates multivariate standard Normal distribution.