arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1710.09718cs.LG
收藏

学习近似随机转移模型

Learning Approximate Stochastic Transition Models

Beijing University of Aeronautics and Astronautics(北京航空航天大学);University of Michigan(密歇根大学);Renmin University of China(中国人民大学);Brown University(布朗大学)

Yuhang Song, Christopher Grimm, Xianming Wang, Michael L. Littman

摘要:

我们研究了从状态到状态的学习映射问题,适用于基于模型的强化学习设置,同时推广到新的状态,可以捕获随机转换。我们表明,目前流行的生成对抗网络努力学习这些随机转移模型,但修改其损失函数的结果在这类问题的一个强大的学习算法。

英文摘要:

We examine the problem of learning mappings from state to state, suitable for use in a model-based reinforcement-learning setting, that simultaneously generalize to novel states and can capture stochastic transitions. We show that currently popular generative adversarial networks struggle to learn these stochastic transition models but a modification to their loss functions results in a powerful learning algorithm for this class of problems.

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。

↑