Benchmark multi-model/multi-view models.

Note

This page is a reference documentation. It only explains the function signature, and not how to use it. Please refer to the gallery for the big picture.

mmbench.workflow.cvae.train_cvae(dataset, datasetdir, outdir, general_lat_dims=15, specific_lat_dims=5, beta=4, lambda1=1, lambda2=2, adam_lr=0.0001, n_epochs=1000, host='http://localhost', port=8085)[source]¶

Train a contrastive Variational Auto-Encoder (cVAE).

Parameters

dataset : str

the dataset name: euaims.

datasetdir : str

the path to the dataset associated data.

outdir : str

the destination folder.

general_lat_dims : int, default 15

the number of latent dimensions in the general part of the latent space.

specific_lat_dims : int, default 5

the number of latent dimensions in the specific part of the latent space.

beta : float, default 4

weight of the KL divergence.

lambda1 : float, default 1

weight for the salient disentanglement loss.

lambda2 : float, default 2

weight for the background disentanglement loss.

adam_lr : float, default 2e-3

the initial learning rate in the ADAM optimizer.

n_epochs : int, default 10000

the number of training epochs.

host : str, default ‘http://localhost’

the host on which visdom is launched.

port : int, default 8085

the port on which the visdom server is launched.