DGSAN: Discrete Generative Self-Adversarial Network

TL;DR

DGSAN introduces a novel adversarial framework for generating discrete data without gradient passing, leveraging iterative updates and theoretical guarantees to outperform existing methods in text generation quality and diversity.

Contribution

The paper proposes a new discrete data generation framework that avoids gradient passing, with iterative updates and theoretical support, improving over prior GAN-based approaches.

Findings

Outperforms recent methods in discrete sequence generation

Provides theoretical guarantees for the proposed approach

Demonstrates superior quality and diversity in experiments

Abstract

Although GAN-based methods have received many achievements in the last few years, they have not been entirelysuccessful in generating discrete data. The most crucial challenge of these methods is the difficulty of passing the gradientfrom the discriminator to the generator when the generator outputs are discrete. Despite the fact that several attemptshave been made to alleviate this problem, none of the existing GAN-based methods have improved the performance oftext generation compared with the maximum likelihood approach in terms of both the quality and the diversity. In thispaper, we proposed a new framework for generating discrete data by an adversarial approach in which there is no need topass the gradient to the generator. The proposed method has an iterative manner in which each new generator is definedbased on the last discriminator. It leverages the discreteness of data and the last discriminator to model the real datadistribution implicitly. Moreover, the method is supported with theoretical guarantees, and experimental results generallyshow the superiority of the proposed DGSAN method compared to the other popular or recent methods in generatingdiscrete sequential data.