Adversarial Approach to Domain Adaptation for Reinforcement Learning on Dialog Systems
Pattern Recognition Letters(2019)
摘要
•We introduced an adversarial approach to aid domain adaptation in Deep Q Network.•We devised a calibrator loss to reduce domain shift at feature network level.•We evaluated the architecture in the PyDial environment with three different domain.•Our method out-performs the baseline, but is hindered by domain-inherit properties.
更多查看译文
关键词
Dialog systems,Reinforcement learning,Domain adaptation,Transfer learning,Deep Q Network,Adversarial networks
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络