Skip to content

zilongzheng/CoopNets

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

53 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Cooperative Learning of Energy-Based Model and Latent Variable Model via MCMC Teaching

This repository contains a tensorflow implementation for the paper "Cooperative Learning of Energy-Based Model and Latent Variable Model via MCMC Teaching". (http://www.stat.ucla.edu/~ywu/CoopNets/main.html)

Requirements

Usage

First, download Imagenet-scene dataset and save it to ./data directory:

$ python download.py scene

To train a model with alp dataset:

$ python main.py --category alp --data_dir ./data/scene --output_dir ./output --num_epochs 300 --batch_size 100 --d_lr 0.01 --g_lr 0.0001

synthesized results will be saved in ./output/alp/synthesis

To test generator by synthesizing interpolation results with trained model:

$ python main.py --test --sample_size 144 --category alp --output_dir ./output --ckpt ./output/alp/checkpoints/model.ckpt

testing results will be saved in ./output/alp/test

Results

Results of MIT Place205 dataset

Descriptor result descriptor

Generator result generator

Interpolation result interpolation

Reference

@inproceedings{coopnets,
    author = {Xie, Jianwen and Lu, Yang and Gao, Ruiqi and Wu, Ying Nian},
    title = {Cooperative Learning of Energy-Based Model and Latent Variable Model via MCMC Teaching},
    booktitle = {The 32nd AAAI Conference on Artitifical Intelligence},
    year = {2018}
}

For any questions, please contact Jianwen Xie (jianwen@ucla.edu) and Zilong Zheng (zilongzheng0318@ucla.edu)

About

Cooperative Learning of Energy-Based Model and Latent Variable Model via MCMC Teaching

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages