API Reference¶
Generative Classification¶
Discriminative Classification¶
- class vbll.layers.classification.DiscClassification(*args: Any, **kwargs: Any)[source]¶
Variational Bayesian Disciminative Classification
Parameters¶
- in_featuresint
Number of input features
- out_featuresint
Number of output features
- regularization_weightfloat
Weight on regularization term in ELBO
- parameterizationstr
Parameterization of covariance matrix. Currently supports {‘dense’, ‘diagonal’, ‘lowrank’}
- softmax_boundstr
Bound to use for softmax. Currently supports ‘jensen’
- return_oodbool
Whether to return OOD scores
- prior_scalefloat
Scale of prior covariance matrix
- wishart_scalefloat
Scale of Wishart prior on noise covariance
- doffloat
Degrees of freedom of Wishart prior on noise covariance
- cov_rankint
Rank of low-rank correction used in the LowRankNormal covariance parameterization.