Generated Method Catalog Snapshot¶

Generated provenance: tools/render_method_catalog.py:render_markdown Source artifacts: reports/method_catalog_latest.json Generated date: 2026-11-08

It is a code-backed snapshot used to reduce docs drift in the task-first matrix.

Summary¶

Total methods: 162
Peer methods present: SWAG, BayesianNeuralNetwork, MixtureDensityLoss

By Family¶

Family	Count
`bnn`	7
`calibration_transform`	7
`causal`	2
`censored`	3
`conformal`	18
`constraints`	2
`density_loss`	3
`eiv`	16
`ensemble`	9
`evidential`	1
`expectile`	2
`flow`	5
`flow_map`	1
`flow_matching`	1
`imbalanced_algorithm`	4
`imbalanced_loss`	11
`inference`	1
`likelihood_ratio`	2
`mc_dropout`	1
`mdn`	1
`multivariate`	1
`observation_mapping`	3
`ordinal`	2
`point_loss`	1
`poisson`	4
`poisson_gaussian`	4
`probabilistic_loss`	11
`quantile`	2
`representation_regularizer`	1
`robust_algorithm`	1
`robust_loss`	9
`semi_supervised`	1
`swag`	2
`tabular_network`	1
`target_transform`	5
`test_time`	8
`tictac`	1
`tweedie`	4
`uncertain_gt`	4

By Maturity¶

Maturity	Count
`Core`	7
`Strong`	20
`Available`	135

Method Rows¶

Legend: * suffix means partial support. This snapshot is for discovery and comparison, not absolute ranking.

Name	Family	Maturity	Task Tags	Core Capability Signals	Public Path
`BayesianNeuralNetwork`	`bnn`	`Available`	`epistemic_uq`, `ood`	`epistemic`, `aleatoric`, `decomposition`, `calibration`, `ood`	`torchregress.ensemble.BayesianNeuralNetwork`
`HeteroscedasticBNN`	`bnn`	`Available`	`uq_decomposition`, `ood`, `calibration`	`epistemic`, `aleatoric`, `decomposition`, `calibration`, `ood`	`torchregress.ensemble.HeteroscedasticBNN`
`HeteroscedasticLaplaceRegressor`	`bnn`	`Available`	`laplace_approximation`, `epistemic_uq`, `aleatoric_uq`, `last_layer`	`epistemic`, `aleatoric`, `decomposition`, `calibration`, `ood`	`torchregress.algorithms.HeteroscedasticLaplaceRegressor`
`IVON`	`bnn`	`Available`	`variational_inference`, `optimizer`, `natural_gradient`, `bayesian_learning_rule`	`epistemic`, `calibration`, `ood`	`torchregress.algorithms.IVON`
`NaturalHeteroscedasticHead`	`bnn`	`Available`	`natural_parameterization`, `optimization_stability`	`aleatoric`, `calibration`, `ood`	`torchregress.algorithms.NaturalHeteroscedasticHead`
`NaturalReparamHead`	`bnn`	`Available`	`natural_parameterization`, `optimization_stability`	`aleatoric`, `calibration`, `ood`	`torchregress.algorithms.NaturalReparamHead`
`VIDSRegressor`	`bnn`	`Available`	`distribution_shift`, `variational_inference`, `adaptive_prior`	`epistemic`, `aleatoric`, `calibration`, `ood`	`torchregress.algorithms.VIDSRegressor`
`BinnedLabelShiftEstimator`	`calibration_transform`	`Strong`	`calibration`, `shift_adaptation`, `label_shift`	`calibration`, `ood*`	`torchregress.calibration.BinnedLabelShiftEstimator`
`CalibratedRegressor`	`calibration_transform`	`Available`	`calibration`, `posthoc_calibration`, `prediction_intervals`	`calibration`, `ood*`	`torchregress.estimators.CalibratedRegressor`
`IsotonicMeanCalibrator`	`calibration_transform`	`Available`	`calibration`, `posthoc_calibration`	`calibration`, `ood*`	`torchregress.calibration.IsotonicMeanCalibrator`
`PITCalibrator`	`calibration_transform`	`Available`	`calibration`, `posthoc_calibration`, `distribution_calibration`	`calibration`, `ood*`	`torchregress.calibration.PITCalibrator`
`SemiConformalCalibrator`	`calibration_transform`	`Strong`	`calibration`, `conformal`, `semi_supervised`, `shift_adaptation`	`calibration`, `ood*`	`torchregress.calibration.SemiConformalCalibrator`
`VarianceRestoringCalibrator`	`calibration_transform`	`Available`	`calibration`, `posthoc_calibration`, `distribution_matching`, `attenuation_bias`	`calibration`, `ood*`	`torchregress.calibration.VarianceRestoringCalibrator`
`VarianceTemperatureScaler`	`calibration_transform`	`Available`	`calibration`, `posthoc_calibration`	`calibration`, `ood*`	`torchregress.calibration.VarianceTemperatureScaler`
`dr_ate`	`causal`	`Available`	`causal_inference`, `ate`, `selection_bias`	`calibration`, `ood`	`torchregress.causal.dr_ate`
`dr_cate`	`causal`	`Available`	`causal_inference`, `cate`, `selection_bias`	`calibration`, `ood`	`torchregress.causal.dr_cate`
`AFTLoss`	`censored`	`Available`	`censored`, `interval_censored`, `survival_regression`	`calibration`, `ood`	`torchregress.losses.AFTLoss`
`CensoredGaussianNLLLoss`	`censored`	`Available`	`censored`, `interval_censored`, `survival_regression`	`calibration`, `ood`	`torchregress.losses.CensoredGaussianNLLLoss`
`CensoredQuantileLoss`	`censored`	`Available`	`censored`, `interval_censored`, `non_gaussian`	`calibration`, `ood`	`torchregress.losses.CensoredQuantileLoss`
`CQR`	`conformal`	`Strong`	`coverage_guarantees`, `calibration`, `heteroscedastic_noise`	`calibration`, `ood*`	`torchregress.losses.CQR`
`CTI`	`conformal`	`Available`	`coverage_guarantees`, `calibration`, `distributional`	`calibration`, `ood*`	`torchregress.losses.CTI`
`CVPlus`	`conformal`	`Available`	`coverage_guarantees`, `calibration`, `ensemble`	`calibration`, `ood*`	`torchregress.losses.CVPlus`
`ConformalLoss`	`conformal`	`Core`	`coverage_guarantees`, `calibration`	`calibration`, `ood*`	`torchregress.losses.ConformalLoss`
`ConformalRegressor`	`conformal`	`Available`	`coverage_guarantees`, `prediction_intervals`, `calibration`	`calibration`, `ood*`	`torchregress.estimators.ConformalRegressor`
`ConformalizedDensity`	`conformal`	`Available`	`conformal`, `density`, `calibration`, `multimodal`	`calibration`, `ood*`	`torchregress.losses.ConformalizedDensity`
`ContrastiveFlowConformal`	`conformal`	`Available`	`conformal`, `contrastive`, `density`, `calibration`	`calibration`, `ood*`	`torchregress.losses.ContrastiveFlowConformal`
`DensityConformal`	`conformal`	`Available`	`coverage_guarantees`, `density_conformal`, `imbalance`	`calibration`, `ood*`	`torchregress.losses.DensityConformal`
`EnsembleBatchCP`	`conformal`	`Available`	`coverage_guarantees`, `calibration`, `ensemble`	`calibration`, `ood*`	`torchregress.losses.EnsembleBatchCP`
`MonteCarloConformal`	`conformal`	`Available`	`coverage_guarantees`, `density_conformal`, `epistemic_uq`	`epistemic`, `calibration`, `ood`	`torchregress.losses.MonteCarloConformal`
`PrevalenceAdjustedCP`	`conformal`	`Available`	`coverage_guarantees`, `density_conformal`, `imbalance`	`calibration`, `ood*`	`torchregress.losses.PrevalenceAdjustedCP`
`SLSConformal`	`conformal`	`Available`	`coverage_guarantees`, `multimodal`, `calibration`	`calibration`, `ood*`	`torchregress.losses.SLSConformal`
`SplitConformal`	`conformal`	`Core`	`coverage_guarantees`, `calibration`, `baseline`	`calibration`, `ood*`	`torchregress.losses.SplitConformal`
`UACQR`	`conformal`	`Available`	`coverage_guarantees`, `calibration`, `heteroscedastic_noise`	`calibration`, `ood*`	`torchregress.losses.UACQR`
`WaldoConformal`	`conformal`	`Available`	`conformal`, `calibration`, `confidence_regions`, `multi_target`	`calibration`, `ood*`	`torchregress.losses.WaldoConformal`
`WaldoCriticalValueRegressor`	`conformal`	`Available`	`conformal`, `calibration`, `confidence_regions`, `heteroscedastic`	`calibration`, `ood*`	`torchregress.losses.WaldoCriticalValueRegressor`
`WaldoDensityConformal`	`conformal`	`Available`	`conformal`, `calibration`, `conditional_density`, `confidence_regions`	`calibration`, `ood*`	`torchregress.losses.WaldoDensityConformal`
`WaldoPosteriorCalibrator`	`conformal`	`Available`	`conformal`, `calibration`, `posterior`, `confidence_regions`	`calibration`, `ood*`	`torchregress.losses.WaldoPosteriorCalibrator`
`BoundedHead`	`constraints`	`Available`	`constraints`, `bounded_outputs`	`calibration`, `ood`	`torchregress.constraints.BoundedHead`
`NonCrossingSort`	`constraints`	`Available`	`constraints`, `ordered_outputs`	`calibration`, `ood`	`torchregress.constraints.NonCrossingSort`
`BinnedDensityLoss`	`density_loss`	`Available`	`scalar_targets`, `conditional_density`, `soft_targets`	`aleatoric`, `calibration`, `ood`	`torchregress.losses.BinnedDensityLoss`
`JointBinnedDensityLoss`	`density_loss`	`Available`	`multi_target`, `conditional_density`, `soft_targets`, `low_dimensional`	`aleatoric`, `calibration`, `ood`	`torchregress.losses.JointBinnedDensityLoss`
`NoiseAwareDensityLoss`	`density_loss`	`Available`	`noisy_labels`, `conditional_density`, `measurement_model`	`aleatoric`, `calibration`, `ood`	`torchregress.losses.NoiseAwareDensityLoss`
`CorrectedScoreEIVLoss`	`eiv`	`Available`	`noisy_features`, `measurement_error`, `attenuation_bias`, `corrected_score`	`calibration`, `ood`	`torchregress.losses.CorrectedScoreEIVLoss`
`ErrorAwareFeatureEncoder`	`eiv`	`Available`	`noisy_features`, `measurement_error`, `feature_encoding`	`calibration`, `ood`	`torchregress.algorithms.ErrorAwareFeatureEncoder`
`ForwardInversionRegressor`	`eiv`	`Available`	`noisy_features`, `measurement_error`, `attenuation_bias`, `generative_model`	`calibration`, `ood`	`torchregress.algorithms.ForwardInversionRegressor`
`FunctionalEIVLoss`	`eiv`	`Available`	`noisy_features`, `measurement_error`	`calibration`, `ood`	`torchregress.losses.FunctionalEIVLoss`
`InputNoiseAugmentationLoss`	`eiv`	`Strong`	`noisy_features`, `measurement_error`, `augmentation`	`calibration`, `ood`	`torchregress.losses.InputNoiseAugmentationLoss`
`InputNoiseBinnedPDFLoss`	`eiv`	`Available`	`noisy_features`, `measurement_error`, `multimodal`	`calibration`, `ood`	`torchregress.losses.InputNoiseBinnedPDFLoss`
`InputNoiseMDNLoss`	`eiv`	`Available`	`noisy_features`, `measurement_error`, `multimodal`	`calibration`, `ood`	`torchregress.losses.InputNoiseMDNLoss`
`LatentMarginalizationLoss`	`eiv`	`Strong`	`noisy_features`, `measurement_error`, `marginalization`	`calibration`, `ood`	`torchregress.losses.LatentMarginalizationLoss`
`LatentNN`	`eiv`	`Available`	`noisy_features`, `measurement_error`, `latent_input`	`calibration`, `ood`	`torchregress.algorithms.LatentNN`
`NoiseAwareRegressor`	`eiv`	`Available`	`noisy_features`, `measurement_error`, `feature_encoding`	`calibration`, `ood`	`torchregress.algorithms.NoiseAwareRegressor`
`NoisyInputPredictor`	`eiv`	`Strong`	`noisy_features`, `measurement_error`, `inference`	`calibration`, `ood`	`torchregress.losses.NoisyInputPredictor`
`OrthogonalDistanceRegressionLoss`	`eiv`	`Available`	`noisy_features`, `measurement_error`	`calibration`, `ood`	`torchregress.losses.OrthogonalDistanceRegressionLoss`
`RegressionCalibration`	`eiv`	`Strong`	`noisy_features`, `measurement_error`, `calibration`	`calibration`, `ood*`	`torchregress.algorithms.RegressionCalibration`
`SIMEX`	`eiv`	`Strong`	`noisy_features`, `measurement_error`, `simex`	`calibration`, `ood`	`torchregress.algorithms.SIMEX`
`SplitMeasurementLoss`	`eiv`	`Available`	`noisy_features`, `measurement_error`, `attenuation_bias`, `repeated_measures`	`calibration`, `ood`	`torchregress.losses.SplitMeasurementLoss`
`StructuralEIVLoss`	`eiv`	`Available`	`noisy_features`, `measurement_error`	`calibration`, `ood`	`torchregress.losses.StructuralEIVLoss`
`BaseEnsembleModel`	`ensemble`	`Core`	`epistemic_uq`, `ood`, `selective_prediction`	`epistemic`, `aleatoric`, `decomposition`, `calibration*`, `ood`	`torchregress.ensemble.BaseEnsembleModel`
`BatchEnsembleRegressor`	`ensemble`	`Available`	`uq_decomposition`, `epistemic_uq`, `aleatoric_uq`, `ood`, `low_compute`	`epistemic`, `aleatoric`, `decomposition`, `calibration`, `ood`	`torchregress.ensemble.BatchEnsembleRegressor`
`BinnedPDFEnsembleModel`	`ensemble`	`Available`	`multimodal_targets`, `non_gaussian`, `calibration`	`epistemic`, `aleatoric`, `decomposition`, `calibration`, `ood`	`torchregress.ensemble.BinnedPDFEnsembleModel`
`CumulativeLinkEnsembleModel`	`ensemble`	`Available`	`ordinal`, `non_gaussian`, `calibration`	`epistemic`, `aleatoric`, `decomposition`, `calibration`, `ood`	`torchregress.ensemble.CumulativeLinkEnsembleModel`
`DensityEnsemble`	`ensemble`	`Available`	`multimodal_targets`, `multi_target`, `conditional_density`, `uq_decomposition`	`epistemic`, `aleatoric`, `decomposition`, `calibration`, `ood`	`torchregress.ensemble.DensityEnsemble`
`HeteroscedasticBatchEnsembleModel`	`ensemble`	`Strong`	`uq_decomposition`, `epistemic_uq`, `aleatoric_uq`, `ood`, `low_compute`	`epistemic`, `aleatoric`, `decomposition`, `calibration`, `ood`	`torchregress.ensemble.HeteroscedasticBatchEnsembleModel`
`HeteroscedasticEnsembleModel`	`ensemble`	`Strong`	`uq_decomposition`, `ood`, `calibration`	`epistemic`, `aleatoric`, `decomposition`, `calibration*`, `ood`	`torchregress.ensemble.HeteroscedasticEnsembleModel`
`MDNEnsembleModel`	`ensemble`	`Available`	`multimodal_targets`, `non_gaussian`, `calibration`, `uq_decomposition`	`epistemic`, `aleatoric`, `decomposition`, `calibration`, `ood`	`torchregress.ensemble.MDNEnsembleModel`
`calibrated_deep_ensemble`	`ensemble`	`Available`	`uq_decomposition`, `epistemic_uq`, `aleatoric_uq`, `calibration`, `tabular`	`epistemic`, `aleatoric`, `decomposition`, `calibration`, `ood`	`torchregress.estimators.calibrated_deep_ensemble`
`EvidentialRegressionLoss`	`evidential`	`Available`	`uq_decomposition`, `single_pass`, `evidential`	`epistemic`, `aleatoric`, `decomposition`, `calibration`, `ood`	`torchregress.losses.EvidentialRegressionLoss`
`ExpectileLoss`	`expectile`	`Available`	`prediction_intervals`, `non_gaussian`, `expectile`	`calibration`, `ood`	`torchregress.losses.ExpectileLoss`
`MultiExpectileLoss`	`expectile`	`Available`	`prediction_intervals`, `non_gaussian`, `multi_expectile`	`calibration`, `ood`	`torchregress.losses.MultiExpectileLoss`
`ContrastiveFlowLoss`	`flow`	`Available`	`parameter_estimation`, `domain_shift`, `simulation_based_inference`	`aleatoric`, `decomposition`, `calibration`, `ood*`	`torchregress.losses.ContrastiveFlowLoss`
`GaussianCopulaSplineHead`	`flow`	`Available`	`multi_target`, `correlated_targets`, `missing_labels`, `conditional_density`	`aleatoric`, `calibration`, `ood`	`torchregress.models.GaussianCopulaSplineHead`
`MarginalContrastiveDensity`	`flow`	`Available`	`multi_target`, `correlated_targets`, `conditional_density`	`aleatoric`, `calibration`, `ood`	`torchregress.models.MarginalContrastiveDensity`
`NormalizingFlowLoss`	`flow`	`Available`	`multimodal`, `non_gaussian`, `multi_target`	`aleatoric`, `decomposition`, `calibration`, `ood*`	`torchregress.losses.NormalizingFlowLoss`
`SLSLoss`	`flow`	`Available`	`multimodal`, `non_gaussian`, `prediction_intervals`	`calibration`, `ood`	`torchregress.losses.SLSLoss`
`NormalizingFlowMap`	`flow_map`	`Available`	`multimodal`, `multi_target`, `conditional_density`, `flow_map`	`aleatoric`, `calibration`, `ood`	`torchregress.models.NormalizingFlowMap`
`FlowMatchingMLP`	`flow_matching`	`Available`	`multimodal`, `multi_target`, `conditional_density`, `sampling`	`aleatoric`, `calibration`, `ood*`	`torchregress.models.FlowMatchingMLP`
`ContinuousGroupDRO`	`imbalanced_algorithm`	`Available`	`imbalance`, `rare_targets`, `group_robust`, `distributionally_robust`	`calibration`, `ood`	`torchregress.algorithms.ContinuousGroupDRO`
`GroupDROTrainer`	`imbalanced_algorithm`	`Available`	`imbalance`, `rare_targets`, `group_robust`, `distributionally_robust`	`calibration`, `ood`	`torchregress.algorithms.GroupDROTrainer`
`VIRLoss`	`imbalanced_algorithm`	`Available`	`imbalance`, `rare_targets`, `imbalanced_regression`, `aleatoric_uq`, `evidential`	`calibration`, `ood`	`torchregress.algorithms.VIRLoss`
`VIRRegressor`	`imbalanced_algorithm`	`Available`	`imbalance`, `rare_targets`, `imbalanced_regression`, `aleatoric_uq`, `evidential`	`calibration`, `ood`	`torchregress.algorithms.VIRRegressor`
`BalancedMSEBMC`	`imbalanced_loss`	`Available`	`imbalance`, `rare_targets`, `long_tail`, `noise_adaptive`	`calibration`, `ood`	`torchregress.losses.BalancedMSEBMC`
`BalancedMSELoss`	`imbalanced_loss`	`Available`	`imbalance`, `rare_targets`, `long_tail`	`calibration`, `ood`	`torchregress.losses.BalancedMSELoss`
`BinReweightedMSELoss`	`imbalanced_loss`	`Available`	`imbalance`, `rare_targets`, `long_tail`	`calibration`, `ood`	`torchregress.losses.BinReweightedMSELoss`
`ConR`	`imbalanced_loss`	`Available`	`imbalance`, `rare_targets`, `representation_learning`, `contrastive`	`calibration`, `ood`	`torchregress.losses.ConR`
`DensityWeightedLoss`	`imbalanced_loss`	`Available`	`imbalance`, `rare_targets`	`calibration`, `ood`	`torchregress.losses.DensityWeightedLoss`
`FeatureDistributionSmoother`	`imbalanced_loss`	`Available`	`imbalance`, `rare_targets`	`calibration`, `ood`	`torchregress.losses.FeatureDistributionSmoother`
`FocalRLoss`	`imbalanced_loss`	`Available`	`imbalance`, `rare_targets`, `focal`	`calibration`, `ood`	`torchregress.losses.FocalRLoss`
`LDSLoss`	`imbalanced_loss`	`Available`	`imbalance`, `rare_targets`	`calibration`, `ood`	`torchregress.losses.LDSLoss`
`PrimeLoss`	`imbalanced_loss`	`Available`	`imbalance`, `rare_targets`, `representation_learning`, `imbalanced_regression`	`calibration`, `ood`	`torchregress.losses.PrimeLoss`
`PropensityWeightedLoss`	`imbalanced_loss`	`Available`	`imbalance`, `selection_bias`, `missing_labels`	`calibration`, `ood`	`torchregress.losses.PropensityWeightedLoss`
`RankSimLoss`	`imbalanced_loss`	`Available`	`imbalance`, `rare_targets`, `representation_learning`, `ranking`	`calibration`, `ood`	`torchregress.losses.RankSimLoss`
`PredictionPoweredInference`	`inference`	`Available`	`inference`, `limited_labels`, `population_estimation`	`calibration`, `ood`	`torchregress.inference.ppi_mean_ci`
`NeuralRatioLoss`	`likelihood_ratio`	`Available`	`parameter_estimation`, `simulation_based_inference`, `domain_shift`, `likelihood_ratio`	`calibration`, `ood`	`torchregress.losses.NeuralRatioLoss`
`WifiRatioEnsemble`	`likelihood_ratio`	`Available`	`parameter_estimation`, `simulation_based_inference`, `likelihood_ratio`, `confidence_intervals`	`epistemic`, `calibration`, `ood`	`torchregress.inference.WifiRatioEnsemble`
`MCDropoutWrapper`	`mc_dropout`	`Strong`	`epistemic_uq`, `low_compute`	`epistemic`, `aleatoric`, `decomposition`, `calibration`, `ood`	`torchregress.ensemble.MCDropoutWrapper`
`MixtureDensityLoss`	`mdn`	`Available`	`multimodal`, `non_gaussian`, `multi_target`	`aleatoric`, `calibration`, `ood`	`torchregress.losses.MixtureDensityLoss`
`TaskAgnosticCorrelations`	`multivariate`	`Available`	`covariance_evaluation`, `conditioning`	`calibration`, `ood`	`torchregress.metrics.TaskAgnosticCorrelations`
`CommonScaleSplineHead`	`observation_mapping`	`Available`	`multi_source`, `common_bias`	`calibration`, `ood`	`torchregress.calibration.CommonScaleSplineHead`
`ConditionBiasHead`	`observation_mapping`	`Available`	`multi_source`, `conditional_bias`	`calibration`, `ood`	`torchregress.calibration.ConditionBiasHead`
`RelativeSurveySplineHead`	`observation_mapping`	`Available`	`multi_source`, `relative_bias`, `correlated_targets`	`calibration`, `ood`	`torchregress.calibration.RelativeSurveySplineHead`
`CumulativeLinkLoss`	`ordinal`	`Available`	`ordinal`, `ordered_targets`, `calibration`	`calibration`, `ood`	`torchregress.losses.CumulativeLinkLoss`
`OrdinalCrossEntropyLoss`	`ordinal`	`Available`	`ordinal`, `ordered_targets`	`calibration`, `ood`	`torchregress.losses.OrdinalCrossEntropyLoss`
`WeightedMSELoss`	`point_loss`	`Core`	`baseline`	`calibration`, `ood`	`torchregress.losses.WeightedMSELoss`
`NegativeBinomialNLLLoss`	`poisson`	`Available`	`count_data`, `non_gaussian`, `overdispersion`	`calibration`, `ood`	`torchregress.losses.NegativeBinomialNLLLoss`
`PoissonDevianceLoss`	`poisson`	`Available`	`count_data`, `non_gaussian`, `poisson`	`calibration`, `ood`	`torchregress.losses.PoissonDevianceLoss`
`PoissonLikelihoodRatioLoss`	`poisson`	`Available`	`count_data`, `non_gaussian`, `poisson`, `ablation`	`calibration`, `ood`	`torchregress.losses.PoissonLikelihoodRatioLoss`
`ZeroInflatedPoissonNLLLoss`	`poisson`	`Available`	`count_data`, `non_gaussian`, `zero_inflated`	`calibration`, `ood`	`torchregress.losses.ZeroInflatedPoissonNLLLoss`
`EnhancedPoissonGaussianContaminationMixtureLoss`	`poisson_gaussian`	`Available`	`count_data`, `non_gaussian`, `mixture`, `noisy_counts`	`calibration`, `ood`	`torchregress.losses.EnhancedPoissonGaussianContaminationMixtureLoss`
`PoissonGaussianContaminationMixtureLoss`	`poisson_gaussian`	`Available`	`count_data`, `non_gaussian`, `mixture`, `noisy_counts`	`calibration`, `ood`	`torchregress.losses.PoissonGaussianContaminationMixtureLoss`
`PoissonGaussianLikelihoodRatioLoss`	`poisson_gaussian`	`Available`	`count_data`, `non_gaussian`, `mixture`, `ablation`	`calibration`, `ood`	`torchregress.losses.PoissonGaussianLikelihoodRatioLoss`
`PoissonGaussianLoss`	`poisson_gaussian`	`Available`	`count_data`, `non_gaussian`, `noisy_counts`	`calibration`, `ood`	`torchregress.losses.PoissonGaussianLoss`
`BetaNLLLoss`	`probabilistic_loss`	`Available`	`heteroscedastic_noise`, `aleatoric_uq`, `optimization_stability`	`aleatoric`, `calibration`, `ood`	`torchregress.losses.BetaNLLLoss`
`FaithfulGaussianLoss`	`probabilistic_loss`	`Available`	`heteroscedastic_noise`, `aleatoric_uq`, `mean_variance_decoupling`	`aleatoric`, `calibration`, `ood`	`torchregress.losses.FaithfulGaussianLoss`
`GaussianCRPSLoss`	`probabilistic_loss`	`Strong`	`heteroscedastic_noise`, `aleatoric_uq`, `calibration`	`aleatoric`, `calibration`, `ood`	`torchregress.losses.GaussianCRPSLoss`
`GaussianLatentForwardObservationNLLLoss`	`probabilistic_loss`	`Available`	`forward_observation`, `measurement_noise`, `quadrature`	`aleatoric`, `calibration`, `ood`	`torchregress.losses.GaussianLatentForwardObservationNLLLoss`
`GaussianNLLLoss`	`probabilistic_loss`	`Core`	`heteroscedastic_noise`, `aleatoric_uq`	`aleatoric`, `calibration`, `ood`	`torchregress.losses.GaussianNLLLoss`
`GaussianWassersteinBoundLoss`	`probabilistic_loss`	`Available`	`heteroscedastic_noise`, `multivariate_targets`, `mean_covariance_supervision`	`aleatoric`, `calibration`, `ood*`	`torchregress.losses.GaussianWassersteinBoundLoss`
`LowRankGaussianLoss`	`probabilistic_loss`	`Strong`	`multi_target`, `scalability`	`aleatoric`, `calibration`, `ood`	`torchregress.losses.LowRankGaussianLoss`
`MarginalizedMultiSourceGaussianNLLLoss`	`probabilistic_loss`	`Available`	`multi_target`, `correlated_targets`, `missing_labels`, `noisy_labels`	`aleatoric`, `calibration`, `ood`	`torchregress.losses.MarginalizedMultiSourceGaussianNLLLoss`
`MultivariateGaussianLoss`	`probabilistic_loss`	`Strong`	`multi_target`, `correlated_targets`	`aleatoric`, `calibration`, `ood`	`torchregress.losses.MultivariateGaussianLoss`
`NeighborhoodCovariancePseudoLabeler`	`probabilistic_loss`	`Available`	`pseudo_labels`, `heteroscedastic_noise`, `self_supervision`	`aleatoric`, `calibration`, `ood*`	`torchregress.algorithms.NeighborhoodCovariancePseudoLabeler`
`RobustMultiSourceGaussianNLLLoss`	`probabilistic_loss`	`Available`	`multi_target`, `correlated_targets`, `missing_labels`, `robust`	`aleatoric`, `calibration`, `ood`	`torchregress.losses.RobustMultiSourceGaussianNLLLoss`
`MultiQuantileLoss`	`quantile`	`Strong`	`prediction_intervals`, `non_gaussian`, `multi_quantile`	`calibration`, `ood*`	`torchregress.losses.MultiQuantileLoss`
`QuantileLoss`	`quantile`	`Core`	`prediction_intervals`, `non_gaussian`	`calibration`, `ood*`	`torchregress.losses.QuantileLoss`
`UncertaintyAwareConR`	`representation_regularizer`	`Available`	`imbalanced_regression`, `noisy_labels`, `scalar_targets`	`ood*`	`torchregress.losses.UncertaintyAwareConR`
`iteratively_reweighted_fit`	`robust_algorithm`	`Strong`	`outliers`, `noisy_labels`, `robust_fitting`, `irls`	`calibration`, `ood`	`torchregress.algorithms.iteratively_reweighted_fit`
`AdaptiveRobustLoss`	`robust_loss`	`Available`	`outliers`, `adaptive_shape`, `learnable`	`calibration`, `ood`	`torchregress.losses.AdaptiveRobustLoss`
`BarronLoss`	`robust_loss`	`Available`	`outliers`, `adaptive_shape`	`calibration`, `ood`	`torchregress.losses.BarronLoss`
`CVaRLoss`	`robust_loss`	`Available`	`outliers`, `tail_focused`, `worst_case`, `noisy_labels`	`calibration`, `ood`	`torchregress.losses.CVaRLoss`
`CauchyLoss`	`robust_loss`	`Available`	`outliers`, `redescending`, `noisy_labels`	`calibration`, `ood`	`torchregress.losses.CauchyLoss`
`CharbonnierLoss`	`robust_loss`	`Strong`	`outliers`, `noisy_labels`, `smooth_approx`	`calibration`, `ood`	`torchregress.losses.CharbonnierLoss`
`LogCoshLoss`	`robust_loss`	`Strong`	`outliers`, `noisy_labels`, `smooth_approx`	`calibration`, `ood`	`torchregress.losses.LogCoshLoss`
`PseudoHuberLoss`	`robust_loss`	`Strong`	`outliers`, `noisy_labels`, `smooth_approx`	`calibration`, `ood`	`torchregress.losses.PseudoHuberLoss`
`TukeyBiweightLoss`	`robust_loss`	`Available`	`outliers`, `redescending`, `noisy_labels`	`calibration`, `ood`	`torchregress.losses.TukeyBiweightLoss`
`WeightedHuberLoss`	`robust_loss`	`Core`	`outliers`, `noisy_labels`	`calibration`, `ood`	`torchregress.losses.WeightedHuberLoss`
`TeacherStudentTrainer`	`semi_supervised`	`Strong`	`semi_supervised`, `weak_supervision`, `pseudo_labeling`	`calibration`, `ood`	`torchregress.semi_supervised.TeacherStudentTrainer`
`MultiSWAG`	`swag`	`Available`	`epistemic_uq`, `ood`	`epistemic`, `aleatoric`, `decomposition`, `calibration`, `ood`	`torchregress.ensemble.MultiSWAG`
`SWAG`	`swag`	`Available`	`epistemic_uq`, `ood`	`epistemic`, `aleatoric`, `decomposition`, `calibration`, `ood`	`torchregress.ensemble.SWAG`
`TabularMLP`	`tabular_network`	`Available`	`baseline`, `tabular`, `heteroscedastic`	`aleatoric`, `calibration`, `ood*`	`torchregress.models.TabularMLP`
`BoxCoxTransformLoss`	`target_transform`	`Available`	`target_transform`, `skewed_targets`	`calibration`, `ood`	`torchregress.losses.BoxCoxTransformLoss`
`LogTransformLoss`	`target_transform`	`Available`	`target_transform`, `skewed_targets`	`calibration`, `ood`	`torchregress.losses.LogTransformLoss`
`SqrtTransformLoss`	`target_transform`	`Available`	`target_transform`, `skewed_targets`	`calibration`, `ood`	`torchregress.losses.SqrtTransformLoss`
`TransformedTargetLoss`	`target_transform`	`Available`	`target_transform`, `skewed_targets`	`calibration`, `ood`	`torchregress.losses.TransformedTargetLoss`
`YeoJohnsonTransformLoss`	`target_transform`	`Available`	`target_transform`, `skewed_targets`	`calibration`, `ood`	`torchregress.losses.YeoJohnsonTransformLoss`
`BayesianLinearHead`	`test_time`	`Available`	`low_shot`, `test_time_adaptation`, `last_layer`, `linear_head`	`epistemic`, `aleatoric`, `decomposition`, `calibration`	`torchregress.test_time.BayesianLinearHead`
`DelayedLabelResidualAdapter`	`test_time`	`Available`	`streaming`, `point_correction`, `variance_inflation`, `cosa`	`calibration`, `ood`	`torchregress.test_time.DelayedLabelResidualAdapter`
`OptimalTransportCoverageGap`	`test_time`	`Available`	`distribution_shift`, `conformal`, `calibration`	`calibration`, `ood`	`torchregress.test_time.OptimalTransportCoverageGap`
`RecursiveBayesianHead`	`test_time`	`Available`	`low_shot`, `test_time_adaptation`, `last_layer`, `streaming`	`epistemic`, `aleatoric`, `decomposition`, `calibration`	`torchregress.test_time.RecursiveBayesianHead`
`ScoreCDFReweighter`	`test_time`	`Available`	`distribution_shift`, `conformal`, `calibration`	`calibration`, `ood`	`torchregress.test_time.ScoreCDFReweighter`
`WeightedConformalRegressionAdapter`	`test_time`	`Available`	`distribution_shift`, `conformal`, `calibration`, `covariate_shift`	`calibration`, `ood*`	`torchregress.test_time.WeightedConformalRegressionAdapter`
`WeightedSplitConformalAdapter`	`test_time`	`Available`	`distribution_shift`, `conformal`, `calibration`	`calibration`, `ood`	`torchregress.test_time.WeightedSplitConformalAdapter`
`weighted_split_classification_predictive_batch`	`test_time`	`Available`	`distribution_shift`, `conformal`, `calibration`	`calibration`, `ood`	`torchregress.test_time.weighted_split_classification_predictive_batch`
`TaylorInducedCovarianceHead`	`tictac`	`Available`	`heteroscedastic_noise`, `multivariate_targets`, `gradient_curvature_parameterization`	`aleatoric`, `calibration`, `ood`	`torchregress.algorithms.TaylorInducedCovarianceHead`
`CompoundPoissonLoss`	`tweedie`	`Available`	`tweedie`, `non_gaussian`, `zero_inflated`, `positive_targets`	`calibration`, `ood`	`torchregress.losses.CompoundPoissonLoss`
`GammaLoss`	`tweedie`	`Available`	`tweedie`, `non_gaussian`, `positive_targets`, `right_skewed`	`calibration`, `ood`	`torchregress.losses.GammaLoss`
`InverseGaussianLoss`	`tweedie`	`Available`	`tweedie`, `non_gaussian`, `positive_targets`	`calibration`, `ood`	`torchregress.losses.InverseGaussianLoss`
`TweedieLoss`	`tweedie`	`Available`	`tweedie`, `non_gaussian`, `positive_targets`, `compound_poisson`	`calibration`, `ood`	`torchregress.losses.TweedieLoss`
`ConsistencyRegLoss`	`uncertain_gt`	`Available`	`uncertain_ground_truth`, `weak_supervision`, `noisy_labels`	`calibration`, `ood`	`torchregress.losses.ConsistencyRegLoss`
`NoisyTargetGaussianNLL`	`uncertain_gt`	`Available`	`uncertain_ground_truth`, `noisy_labels`, `weak_supervision`	`aleatoric`, `calibration`, `ood`	`torchregress.losses.NoisyTargetGaussianNLL`
`PseudoLabelConsistencyLoss`	`uncertain_gt`	`Available`	`uncertain_ground_truth`, `weak_supervision`, `semi_supervised`	`calibration`, `ood`	`torchregress.losses.PseudoLabelConsistencyLoss`
`PseudoLabelNLL`	`uncertain_gt`	`Available`	`uncertain_ground_truth`, `weak_supervision`, `noisy_labels`	`calibration`, `ood`	`torchregress.losses.PseudoLabelNLL`

Capability-Focused Shortlists¶

Multimodal (`multimodal=yes`)¶

SLSConformal, BinnedDensityLoss, JointBinnedDensityLoss, InputNoiseBinnedPDFLoss, InputNoiseMDNLoss, DensityEnsemble, ContrastiveFlowLoss, NormalizingFlowLoss, SLSLoss, NormalizingFlowMap, FlowMatchingMLP, MixtureDensityLoss, EnhancedPoissonGaussianContaminationMixtureLoss, PoissonGaussianContaminationMixtureLoss

Noisy Features / EIV (`noisy_features_eiv=yes`)¶

CorrectedScoreEIVLoss, ErrorAwareFeatureEncoder, ForwardInversionRegressor, FunctionalEIVLoss, InputNoiseAugmentationLoss, InputNoiseBinnedPDFLoss, InputNoiseMDNLoss, LatentMarginalizationLoss, LatentNN, NoiseAwareRegressor, NoisyInputPredictor, OrthogonalDistanceRegressionLoss, RegressionCalibration, SIMEX, SplitMeasurementLoss, StructuralEIVLoss

Decomposition (`decomposition=yes`)¶

HeteroscedasticBNN, HeteroscedasticLaplaceRegressor, DensityEnsemble, HeteroscedasticBatchEnsembleModel, HeteroscedasticEnsembleModel, MDNEnsembleModel, EvidentialRegressionLoss