BriefGPT.xyz
Aug, 2019
ARAML:一个文本生成的稳定对抗训练框架
ARAML: A Stable Adversarial Training Framework for Text Generation
HTML
PDF
Pei Ke, Fei Huang, Minlie Huang, Xiaoyan Zhu
TL;DR
提出了一种称为ARAML的新框架,通过采用最大似然估计算法以及与强化学习中不同的奖励函数,来优化文本GAN模型的性能。实验证明,这种模型能比现有模型更稳定、更优秀地生成文本。
Abstract
Most of the existing
generative adversarial networks
(GAN) for
text generation
suffer from the instability of
reinforcement learning
train
→