Maximum-Likelihood Augmented Discrete Generative Adversarial Networks
This addresses a fundamental problem in machine learning for researchers and practitioners working with discrete data like text, offering a more stable training method for GANs in such domains.
The paper tackles the challenge of applying generative adversarial networks (GANs) to discrete settings, such as natural language tasks, by proposing a novel low-variance objective derived from the discriminator's output that corresponds to log-likelihood, resulting in improved stability and effectiveness across various discrete datasets.
Despite the successes in capturing continuous distributions, the application of generative adversarial networks (GANs) to discrete settings, like natural language tasks, is rather restricted. The fundamental reason is the difficulty of back-propagation through discrete random variables combined with the inherent instability of the GAN training objective. To address these problems, we propose Maximum-Likelihood Augmented Discrete Generative Adversarial Networks. Instead of directly optimizing the GAN objective, we derive a novel and low-variance objective using the discriminator's output that follows corresponds to the log-likelihood. Compared with the original, the new objective is proved to be consistent in theory and beneficial in practice. The experimental results on various discrete datasets demonstrate the effectiveness of the proposed approach.