ConvAEMSPSEncoder

class behavenet.models.vaes.ConvAEMSPSEncoder(hparams)

Bases: behavenet.models.aes.ConvAEEncoder

Convolutional encoder that separates label-related subspace.

Methods Summary

forward(x[, dataset])

Process input data.

Methods Documentation

forward(x, dataset=None)

Process input data.

Parameters

• x (torch.Tensor object) – input data

• dataset (int) – used with session-specific io layers

Returns

• encoder output y (torch.Tensor): constrained latents (predicted labels) of shape (n_labels)

• encoder output z (torch.Tensor): unconstrained latents of shape (n_latents - n_labels)

• logvar (torch.Tensor): log variance of latents of shape (n_latents)

• pool_idx (list): max pooling indices for each layer

• output_size (list): output size for each layer

Return type

tuple