arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1611.09321cs.LGcs.AI
收藏

挖掘未被重视的奖励提升策略梯度

Improving Policy Gradient by Exploring Under-appreciated Rewards

Google Brain(谷歌大脑);University of Alberta(阿尔伯塔大学)

Ofir Nachum, Mohammad Norouzi, Dale Schuurmans

摘要:

针对无模型强化学习(RL)提出了一种新的策略梯度形式,并改进了其探索特性.当前基于策略的方法使用熵正则化来鼓励对奖励景观的无向探索,这在具有稀疏奖励的高维空间中是无效的。我们提出了一个更有针对性的探索策略,促进探索低估的奖励区域。如果一个动作序列在当前策略下的对数概率低估了它的回报,那么它就被认为是被低估了。所提出的探索策略易于实现,需要对REINFORCE算法的实现进行小的修改。我们评估了一组长期挑战RL方法的算法任务的方法。我们的方法降低了超参数的敏感性,并表现出显着的改进基线方法。我们的算法成功地解决了一个基准的多位数加法任务,并推广到长序列。据我们所知,这是第一次纯RL方法仅使用奖励反馈解决加法。

英文摘要:

This paper presents a novel form of policy gradient for model-free reinforcement learning (RL) with improved exploration properties. Current policy-based methods use entropy regularization to encourage undirected exploration of the reward landscape, which is ineffective in high dimensional spaces with sparse rewards. We propose a more directed exploration strategy that promotes exploration of under-appreciated reward regions. An action sequence is considered under-appreciated if its log-probability under the current policy under-estimates its resulting reward. The proposed exploration strategy is easy to implement, requiring small modifications to an implementation of the REINFORCE algorithm. We evaluate the approach on a set of algorithmic tasks that have long challenged RL methods. Our approach reduces hyper-parameter sensitivity and demonstrates significant improvements over baseline methods. Our algorithm successfully solves a benchmark multi-digit addition task and generalizes to long sequences. This is, to our knowledge, the first time that a pure RL method has solved addition using only reward feedback.

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。

↑