Learning Approximate Stochastic Transition Models
Learning Approximate Stochastic Transition Models
复制标题
DOI：
--
复制
发表时间：
2017-10
期刊：
影响因子：
--
通讯作者：
Pichao Wang;W. Li;Jun Wan;P. Ogunbona;Xinwang Liu
中科院分区：
文献类型：
--
作者：
Pichao Wang;W. Li;Jun Wan;P. Ogunbona;Xinwang Liu
我们研究了状态到状态的学习映射问题，适用于基于模型的强化学习环境，同时推广到新的状态并可以捕获随机转移。我们表明，目前流行的生成性对抗网络很难学习这些随机转移模型，但对它们的损失函数的修改可以为这类问题提供一个强大的学习算法。
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.