twodlearn.bayesnet.linear module¶

class twodlearn.bayesnet.linear.VLinear(**kargs)[source]¶

Bases: twodlearn.core.common.TdlModel

variational linear model

Tdl autoinitialization with arguments:

weights[source]¶: (Submodel) linear weights

basis_dim[source]¶: (InputArgument) number of dimensions of the explicit basis

weights_prior[source]¶: (Submodel) prior for the linear parameters

basis[source]¶: (InputModel)

batch_shape[source]¶: (InputArgument) number of output dimensions

class VLinearELBO(**kargs)[source]¶

Bases: twodlearn.core.common.TdlModel

property basis_x[source]¶

batch_size[source]¶

dataset_size[source]¶: number of samples in the entire dataset

expected_log_likelihood[source]¶

property inputs[source]¶

labels[source]¶

melbo[source]¶

posterior[source]¶

value[source]¶

property weights[source]¶

weights_kl[source]¶

property weights_prior[source]¶

property y_scale[source]¶

class VLinearTransform(**kargs)[source]¶

Bases: twodlearn.bayesnet.distributions.MVN

distribution after performing a linear transformation

basis_x[source]¶

covariance[source]¶

inputs[source]¶

loc[source]¶

model[source]¶: VLinear model

neg_melbo(labels, dataset_size=None)[source]¶: return the negative mean elbo

property weights[source]¶

property weights_prior[source]¶

with_noise()[source]¶: return the posterior with noise

y_scale[source]¶: linear operation for y_scale

basis[source]

basis_dim[source]: number of dimensions of the explicit basis

batch_shape[source]: number of output dimensions

weights[source]: linear weights

weights_prior[source]: prior for the linear parameters