SeqGAN: Sequence Generative Adversarial Nets with Policy Gradient

https://arxiv.org/abs/1609.05473

Lantao Yu,  Weinan Zhang,  Jun Wang,  Yong Yu

(Submitted on 18 Sep 2016 ( v1), last revised 9 Dec 2016 (this version, v5))

As a new way of training generative models, Generative Adversarial Nets (GAN) that uses a discriminative model to guide the training of the generative model has enjoyed considerable success in generating real-valued data. However, it has limitations when the goal is for generating sequences of discrete tokens. A major reason lies in that the discrete outputs from the generative model make it difficult to pass the gradient update from the discriminative model to the generative model. Also, the discriminative model can only assess a complete sequence, while for a partially generated sequence, it is non-trivial to balance its current score and the future one once the entire sequence has been generated. In this paper, we propose a sequence generation framework, called SeqGAN, to solve the problems. Modeling the data generator as a stochastic policy in reinforcement learning (RL), SeqGAN bypasses the generator differentiation problem by directly performing gradient policy update. The RL reward signal comes from the GAN discriminator judged on a complete sequence, and is passed back to the intermediate state-action steps using Monte Carlo search. Extensive experiments on synthetic data and real-world tasks demonstrate significant improvements over strong baselines.

Comments:	The Thirty-First AAAI Conference on Artificial Intelligence (AAAI 2017)
Subjects:	Learning (cs.LG); Artificial Intelligence (cs.AI)
Cite as:	arXiv:1609.05473 [cs.LG]
	(or arXiv:1609.05473v5 [cs.LG] for this version)