Denoising Adversairal Autoencoders
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
DAAE_v1_omni_params added folders for saving results Mar 1, 2017
DAAE_v1_params added folders for saving results Mar 1, 2017
DAAE_v1_sprite_params added folders for saving results Mar 1, 2017
DAAE_v2_omni_params added folders for saving results Mar 1, 2017
DAAE_v2_params
DAAE_v2_sprite_params added folders for saving results Mar 1, 2017
InData removed 'InData/sprites/train.npy.tar.gz' Mar 1, 2017
DAAE_omni_v1.ipynb add code added Mar 1, 2017
DAAE_omni_v2.ipynb
DAAE_sprite_v1.ipynb add code added Mar 1, 2017
DAAE_sprite_v2.ipynb DAAE sprite v2 for M=5 Mar 2, 2017
DAAE_v1.ipynb add code added Mar 1, 2017
DAAE_v2.ipynb add code added Mar 1, 2017
PCA_Exp.ipynb add code added Mar 1, 2017
README.md add code added Mar 1, 2017
ll.py add code added Mar 1, 2017
parzen.py add code added Mar 1, 2017
requirements.txt

README.md

DAAE

Denosing Adversairal Autoencoder

This repo contains the code needed to run the experiments from the paper "Denoising Adversarial Autoencoders", Antonia Creswell and Anil Anthony Bharath

We provide examples of experimental results for each of our networks (DAAE and iDAAE) trained on 3 datasets. We do not (yet) provide the trained models but do provide the network and training parameters needed to replicate our results. Note that results, expecially for the log-likelihood may vary because of the stoachstic nature of the image generation process.

#To use the code:

  1. clone the repo
  2. unzip all the compressed data files
  3. install lasagne, theano, matplotlib, scikit-image, numpy, dill, ipython, jupyter, scikit-learn ... (a full list may be found it reqiurements.txt)
  4. run the following from comand line to start a notebook: $ ipython notebook
  5. run the code by pressing the "play" (triangle |>) button