Constantinos Daskalakis, Andrew Ilyas, Vasilis Syrgkanis, Haoyang Zeng
TL;DR本文提出使用乐观镜像下降(OMD)解决在训练生成对抗网络中的极限循环问题,通过实验验证 OMD 解决了 WGAN 训练中的极限循环问题并同时提高了模型性能,尤其应用于生物信息学中生成 DNA 序列的任务中表现优异。
Abstract
We address the issue of limit cycling behavior in training Generative
Adversarial Networks and propose the use of optimistic mirror decent (OMD) for
training wasserstein gans. Recent theoretical results have show