transformation

Transformations for models.

Submodules

batchensemble(base[, num_members, ...])

Create a BatchEnsemble predictor from a base predictor based on [WTB20].

bayesian(base[, use_base_weights, ...])

Create a Bayesian predictor from a base predictor based on [BCKW15].

bayesian_ensemble(base[, use_base_weights, ...])

Create an ensemble of Bayesian-layer transformed predictors based on [BCKW15].

calibration

Calibration transformations for logit predictors.

cast(base)

Return a predictor unchanged while optionally registering its predictor type.

class_bias_ensemble(base, num_members[, ...])

Create an ensemble with class-specific final-layer bias initialization based on [LohrHMHullermeier25].

conformal

Conformal transformations for regression and classification.

conformal_credal_set

Conformalized credal set prediction.

dirichlet_clipped_exp_one_activation(base)

Append clipped exp plus one so logits become Dirichlet concentrations.

dirichlet_exp_activation(base)

Append an exp activation so logits become Dirichlet concentrations.

dirichlet_softplus_activation(base)

Append Softplus plus one so logits become Dirichlet concentrations.

dropconnect(base[, p, rng_collection, rngs])

Create a DropConnect predictor from a base predictor based on [MYM+21].

dropout(base[, p, rng_collection, rngs, ...])

Create a Dropout predictor from a base predictor based on [GG16].

ensemble(base, num_members[, reset_params])

Create an ensemble predictor from a base predictor based on [LPB17].

interval_classifier(base[, use_base_weights])

Create an interval classifier from a base classifier based on [WSM+25].

masksembles(base[, num_masks, scale])

Create a Masksembles predictor from a base predictor based on [DBBaqueF21].

natural_posterior_network

Module for natural posterior network implementations.

normal_inverse_gamma_head(base)

Replace the final linear layer with a Normal-Inverse-Gamma head based on [ASSR20].

posterior_network(encoder, latent_dim, ...)

Create a Posterior Network predictor from an encoder based on [CZugnerGunnemann20].

subensemble(base, num_heads[, head, ...])

Create a subensemble predictor from a base model and optional head based on [VT19].

transformation

Utilities for the definition of methods.

Functions

batchensemble

Create a BatchEnsemble predictor from a base predictor based on [WTB20].

bayesian

Create a Bayesian predictor from a base predictor based on [BCKW15].

bayesian_ensemble

Create an ensemble of Bayesian-layer transformed predictors based on [BCKW15].

cast

Return a predictor unchanged while optionally registering its predictor type.

class_bias_ensemble

Create an ensemble with class-specific final-layer bias initialization based on [LohrHMHullermeier25].

conformal_dirichlet_relative_likelihood

Create a credal predictor with Dirichlet level-set conformal sets.

conformal_inner_product

Create a credal predictor with inner-product-ball conformal sets.

conformal_kullback_leibler

Create a credal predictor with KL-ball conformal sets.

conformal_total_variation

Create a credal predictor with TV-ball conformal sets.

conformal_wasserstein_distance

Create a credal predictor with Wasserstein-ball conformal sets.

dirichlet_clipped_exp_one_activation

Append clipped exp plus one so logits become Dirichlet concentrations.

dirichlet_exp_activation

Append an exp activation so logits become Dirichlet concentrations.

dirichlet_softplus_activation

Append Softplus plus one so logits become Dirichlet concentrations.

dropconnect

Create a DropConnect predictor from a base predictor based on [MYM+21].

dropout

Create a Dropout predictor from a base predictor based on [GG16].

ensemble

Create an ensemble predictor from a base predictor based on [LPB17].

interval_classifier

Create an interval classifier from a base classifier based on [WSM+25].

masksembles

Create a Masksembles predictor from a base predictor based on [DBBaqueF21].

normal_inverse_gamma_head

Replace the final linear layer with a Normal-Inverse-Gamma head based on [ASSR20].

posterior_network

Create a Posterior Network predictor from an encoder based on [CZugnerGunnemann20].

subensemble

Create a subensemble predictor from a base model and optional head based on [VT19].