Generative Adversarial Networks (GANs) have shown great promise recently in image generation. Training GANs for language generation has proven to be more difficult, because of the non-differentiable nature of generating text with recurrent neural networks. Consequently, past work has either resorted to pre-training with maximum-likelihood or used convolutional networks for generation. In this work, we show that recurrent neural networks can be trained to generate text with GANs from scratch using curriculum learning, by slowly teaching the model to generate sequences of increasing and variable length. We empirically show that our approach vastly improves the quality of generated sequences compared to a convolutional baseline.
@article{arxiv.1706.01399,
title = {Language Generation with Recurrent Generative Adversarial Networks without Pre-training},
author = {Ofir Press and Amir Bar and Ben Bogin and Jonathan Berant and Lior Wolf},
journal= {arXiv preprint arXiv:1706.01399},
year = {2017}
}
Comments
Presented at the 1st Workshop on Learning to Generate Natural Language at ICML 2017