arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2009.09467cs.LGcs.ROstat.ML
收藏

用中性奖励函数解决对抗性模仿学习中的奖励偏差

Addressing reward bias in Adversarial Imitation Learning with neutral reward functions

The Robotics Institute(机器人研究所);Carnegie Mellon University(卡内基梅隆大学)

Rohit Jena, Siddharth Agrawal, Katia Sycara

摘要:

生成式对抗模仿学习存在奖励偏差的基本问题,该问题源于算法中使用的奖励函数的选择。不同类型的偏见也会影响不同类型的环境,这些环境大致分为生存环境和基于任务的环境。我们提供了一个理论上的草图,为什么现有的奖励功能会失败的模仿学习的情况下,基于任务的环境中,多个终端状态。我们还提出了一个新的GAIL奖励函数,它优于现有的GAIL方法基于任务的环境中的单个和多个终端状态,并有效地克服了生存和终止偏见。

英文摘要:

Generative Adversarial Imitation Learning suffers from the fundamental problem of reward bias stemming from the choice of reward functions used in the algorithm. Different types of biases also affect different types of environments - which are broadly divided into survival and task-based environments. We provide a theoretical sketch of why existing reward functions would fail in imitation learning scenarios in task based environments with multiple terminal states. We also propose a new reward function for GAIL which outperforms existing GAIL methods on task based environments with single and multiple terminal states and effectively overcomes both survival and termination bias.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑