Skip to content

Latest commit

 

History

History
32 lines (23 loc) · 1005 Bytes

README.md

File metadata and controls

32 lines (23 loc) · 1005 Bytes

Relational-AutoEncoders

This package includes the implementation of my ICML 2020 work "Learning Autoencoders with Relational Regularization" [https://arxiv.org/pdf/2002.02913.pdf]

Main Dependencies

  • argparse
  • matplotlib
  • numpy
  • pickle
  • pytorch
  • sklearn

Platform:

We test this example in a conda environment on Windows 10, with cuda 10.1 and one 1080Ti GPU

Test our method:

  1. Open a terminal and go to the folder of the example.
  2. python test_rae.py --model-type deterministic --source-data DATANAME (Learning a deterministic RAE for a dataset.)
  3. python test_rae.py --model-type probabilistic --source-data DATANAME (Learning a probabilistic RAE for a dataset.)
  • The DATANAME can be MNIST and CelebA

Test baselines:

  1. Open a terminal and go to the folder of the example.
  2. python test_MODEL.py --source-data DATANAME
  • The MODEL can be vae, wae, swae, gmvae, and vampprior

All the results are in the folder "Results".