BriefGPT.xyz
Sep, 2017
ALICE: 面向联合分布匹配的对抗学习理解
Towards Understanding Adversarial Learning for Joint Distribution Matching
HTML
PDF
Chunyuan Li, Hao Liu, Changyou Chen, Yunchen Pu, Liqun Chen...
TL;DR
本研究旨在探究双向对抗训练中的不可识别性问题,并提出了基于条件熵的对抗和非对抗方法,用于无监督和监督任务的匹配联合分布学习。同时,将广泛的对抗模型统一为联合分布匹配问题,并提供了对于半监督学习任务的扩展方法。通过对合成数据和现实世界应用的验证,证明了理论结果的正确性。
Abstract
We investigate the non-identifiability issues associated with
bidirectional adversarial training
for
joint distribution matching
. Within a framework of
→