CosmosShadow / SeqGAN

Implementation of Sequence Generative Adversarial Nets with Policy Gradient

SeqGAN

Requirements:

• Tensorflow r1.0.1
• Python 2.7
• CUDA 7.5+ (For GPU)

Introduction

Apply Generative Adversarial Nets to generating sequences of discrete tokens.

The illustration of SeqGAN.

Left: D is trained over the real data and the generated data by G.

Right: G is trained by policy gradient where the final reward signal is provided by D and is passed back to the intermediate action value via Monte Carlo search.

SeqGAN: Sequence Generative Adversarial Nets with Policy Gradient

train

$ python sequence_gan.py

The experiment has two stages

• use the positive data provided by the oracle model and Maximum Likelihood Estimation to perform supervise learning.
• use adversarial training to improve the generator.