arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2102.10330cs.LGcs.AI
收藏

强化学习中的解耦值与泛化策略

Decoupling Value and Policy for Generalization in Reinforcement Learning

New York University(纽约大学)

Roberta Raileanu, Rob Fergus

摘要:

标准的深度强化学习算法使用策略和价值函数的共享表示,特别是在直接从图像进行训练时。然而,我们认为,需要更多的信息来准确地估计价值函数比学习的最优策略。因此,使用策略和价值函数的共享表示可能导致过拟合。为了缓解这个问题,我们提出了两种方法,结合起来创建IDAAC:不变解耦优势演员批评。首先,IDAAC使用单独的网络对策略和价值函数进行建模,从而实现了策略和价值函数的优化。其次,它引入了一个辅助损失,鼓励表示是不变的任务无关的环境属性。IDAAC对看不见的环境表现出良好的泛化能力,在Procgen基准测试中达到了新的最先进水平,并在有干扰物的DeepMind控制任务中表现优于流行方法。我们的实现可以在https://github.com/rraileanu/idaac上获得。

英文摘要:

Standard deep reinforcement learning algorithms use a shared representation for the policy and value function, especially when training directly from images. However, we argue that more information is needed to accurately estimate the value function than to learn the optimal policy. Consequently, the use of a shared representation for the policy and value function can lead to overfitting. To alleviate this problem, we propose two approaches which are combined to create IDAAC: Invariant Decoupled Advantage Actor-Critic. First, IDAAC decouples the optimization of the policy and value function, using separate networks to model them. Second, it introduces an auxiliary loss which encourages the representation to be invariant to task-irrelevant properties of the environment. IDAAC shows good generalization to unseen environments, achieving a new state-of-the-art on the Procgen benchmark and outperforming popular methods on DeepMind Control tasks with distractors. Our implementation is available at https://github.com/rraileanu/idaac.

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。

↑