Helper Module for Deep Learning.

Sparse Multi-Channel Variational Autoencoderfor the Joint Analysis of Heterogeneous Data.

class pynet.models.vae.mcvae.MCVAE(latent_dim, n_channels, n_feats, noise_init_logvar=-3, noise_fixed=False, sparse=False, vae_model='dense', vae_kwargs=None, nodecoding=False)[source]¶

Sparse Multi-Channel Variational Autoencoder (sMCVAE).

Sparse Multi-Channel Variational Autoencoder for the Joint Analysis of Heterogeneous Data, Luigi Antelmi, Nicholas Ayache, Philippe Robert, Marco Lorenzi Proceedings of the 36th International Conference on Machine Learning, PMLR 97:302-311, 2019.

__init__(latent_dim, n_channels, n_feats, noise_init_logvar=-3, noise_fixed=False, sparse=False, vae_model='dense', vae_kwargs=None, nodecoding=False)[source]¶

Init class.

Parameters

latent_dim: int

the number of latent dimensions.

n_channels: int

the number of channels.

n_feats: list of int

each channel input dimensions.

noise_init_logvar: float, default -3

default noise parameters values.

noise_fixed: bool, default False

if set not set do not required gradients on noise parameters.

sparse: bool, default False

use sparsity contraint.

vae_model: str, default “dense”

the VAE network used to encode each channel.

vae_kwargs: dict, default None

extra parameters passed initialization of the VAE model.

nodecoding: bool, default False

if set do not apply the decoding.

apply_threshold(z, threshold, keep_dims=True, reorder=False)[source]¶

Apply dropout threshold.

Parameters

z: Tensor

distribution samples.

threshold: float

dropout threshold.

keep_dims: bool default True

dropout lower than threshold is set to 0.

reorder: bool default False

reorder dropout rates.

Returns

z_keep: list

dropout rates.

decode(z)[source]¶

Maps the given latent codes onto the image space.

Parameters

z: list of Tensor (N, D)

sample from the distribution having latent parameters mu, var.

Returns

p: list of Tensor, (N, C, F)

the prediction p(x|z).

property dropout¶

encode(x)[source]¶

Encodes the input by passing through the encoder network and returns the latent distribution for each channel.

Parameters

x: list of Tensor, (C,) -> (N, Fc)

input tensors to encode.

Returns

out: list of 2-uplet (C,) -> (N, D)

each channel distribution parameters mu (mean of the latent Gaussian) and logvar (standard deviation of the latent Gaussian).

forward(x)[source]¶

init_vae()[source]¶

Create one VAE model per channel.

p_to_prediction(p)[source]¶

Get the prediction from various types of distributions.

reconstruct(p)[source]¶