/ORGAN

Objective-Reinforced Generative Adversarial Networks (ORGAN) for Sequence Generation Models

Primary LanguagePython

Objective-Reinforced GANs (ORGAN)

  • Want the diversity and interestingness that you get with samples from an adversarial process (GAN)?

  • Want the directed focus you can give algorithms with Reinforcement Learning? (RL)

  • Working with discrete sequence data (text, molecular SMILES, abc musical notation ,etc.)?

Then ORGAN if for you, define simple reward functions and alternate between adversarial and reinforced training.

Based on work from [](arxiv link here)

How to train

In order to train the model, cd into model and run

python train_ogan.py exp.json

where exp.json is a experiment configuration file.

A GPU is recommended since it can take several days to run, depending on dataset and sequence extension, algorithm is not parallelized for multiple GPUs.

How to sample

Requirements to run

  • Tensorflow 1.0
  • Python 2 or 3
  • rdkit for molecular purposes

Make your own experiment

Dockerfile