arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2402.11367cs.LG
收藏

多任务逆强化学习用于常识奖励

Multi Task Inverse Reinforcement Learning for Common Sense Reward

Bar-Ilan University(巴伊兰大学)

Neta Glazer, Aviv Navon, Aviv Shamsian, Ethan Fetaya

AI总结:

本文提出将奖励分解为任务特定与常识两部分,并证明多任务逆强化学习能从专家示范中学习有效的常识奖励,避免奖励黑客问题。

摘要:

在复杂真实环境中应用强化学习的挑战之一在于为智能体提供足够详细的奖励函数。奖励与期望行为之间的任何偏差都可能导致不期望的结果。这可能会引发诸如“奖励黑客”等问题,即智能体通过非预期行为最大化奖励。在这项工作中,我们提出将奖励分解为两个不同的部分:一个简单的任务特定奖励,概述当前任务的具体细节;以及一个未知的常识奖励,指示智能体在环境中的期望行为。然后,我们探讨如何从专家示范中学习这种常识奖励。我们首先表明,逆强化学习,即使成功训练了智能体,也不能学习到有用的奖励函数。也就是说,用学习到的奖励训练新智能体并不会损害期望行为。然后,我们证明这个问题可以通过同时在多个任务上训练来解决。即,多任务逆强化学习可以应用于学习有用的奖励函数。

英文摘要:

One of the challenges in applying reinforcement learning in a complex real-world environment lies in providing the agent with a sufficiently detailed reward function. Any misalignment between the reward and the desired behavior can result in unwanted outcomes. This may lead to issues like "reward hacking" where the agent maximizes rewards by unintended behavior. In this work, we propose to disentangle the reward into two distinct parts. A simple task-specific reward, outlining the particulars of the task at hand, and an unknown common-sense reward, indicating the expected behavior of the agent within the environment. We then explore how this common-sense reward can be learned from expert demonstrations. We first show that inverse reinforcement learning, even when it succeeds in training an agent, does not learn a useful reward function. That is, training a new agent with the learned reward does not impair the desired behaviors. We then demonstrate that this problem can be solved by training simultaneously on multiple tasks. That is, multi-task inverse reinforcement learning can be applied to learn a useful reward function.

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。

↑