Chainer implementation of adversarial autoencoder (AAE)
Python
Latest commit f1e030d Dec 6, 2016 @musyoku rewrite
Permalink
Failed to load latest commit information.
semi-supervised rewrite Dec 6, 2016
sequential rewrite Dec 6, 2016
supervised/learn_style rewrite Dec 6, 2016
unsupervised rewrite Dec 6, 2016
README.md rewrite Dec 6, 2016
aae.py rewrite Dec 6, 2016
aae_dim_reduction.py rewrite Dec 6, 2016
aae_regularized.py rewrite Dec 6, 2016
aae_semi_supervised.py rewrite Dec 6, 2016
aae_supervised.py rewrite Dec 6, 2016
dataset.py rewrite Dec 6, 2016
mnist_tools.py rewrite Dec 6, 2016
params.py rewrite Dec 6, 2016
progress.py rewrite Dec 6, 2016
sampler.py rewrite Dec 6, 2016
visualizer.py rewrite Dec 6, 2016

README.md

Adversarial AutoEncoder

code for the paper

Requirements

  • Chainer 1.18
  • pylab
  • PIL
  • gzip
  • six

Contains the following repository:

Incorporating Label Information in the Adversarial Regularization

run semi-supervised/regularize_z/train.py

We trained with a prior (a mixture of 10 2-D Gaussians or Swissroll distribution) on 10K labeled MNIST examples and 40K unlabeled MNIST examples.

gaussian

swissroll

Supervised Adversarial Autoencoders

run supervised/learn_style/train.py

analogy

Semi-Supervised Adversarial Autoencoders

run semi-supervised/classification/train.py

data #
labeled 100
unlabeled 49900
validation 10000

Validation accuracy for each epoch

classification

Analogies

analogy_semi

Unsupervised Clustering

run unsupervised/clustering/train.py

16 clusters

clusters_16

32 clusters

clusters_32

Dimensionality Reduction

run unsupervised/dim_reduction/train.py

reduction_unsupervised

run semi-supervised/dim_reduction/train.py

reduction_100