人工智能培训

搜索

机器学习论文:用于合作强化学习的封建多智能体层次结构(Feudal Multi-Agent Hierarchies for Cooperative Reinfo

[复制链接]
1449231467 发表于 2019-1-28 12:04:43 | 显示全部楼层 |阅读模式
1449231467 2019-1-28 12:04:43 125 0 显示全部楼层
机器学习论文:用于合作强化学习的封建多智能体层次结构(Feudal Multi-Agent Hierarchies for Cooperative Reinforcement Learning)我们研究强化学习机构如何学会合作。从人类社会中汲取灵感,其中通常通过等级组织来促进多个人的成功协调,我们引入了封建多智能体层次结构(FMH)。在这个框架中,负责最大化环境决定的奖励功能的“经理”代理学习将子目标传达给多个同时操作的“工人”代理。为实现管理子目标而获得奖励的工人在世界上同时采取行动。我们概述了FMH的结构,并展示了其分散学习和控制的潜力。我们发现,给定一组足够的子目标供选择,FMH执行,特别是比例,比使用共享奖励功能的合作方法更好。
We investigate how reinforcement learning agents can learn to cooperate.Drawing inspiration from human societies, in which successful coordination ofmany individuals is often facilitated by hierarchical organisation, weintroduce Feudal Multi-agent Hierarchies (FMH).In this framework, a 'manager'agent, which is tasked with maximising the environmentally-determined rewardfunction, learns to communicate subgoals to multiple, simultaneously-operating,'worker' agents.Workers, which are rewarded for achieving managerial subgoals,take concurrent actions in the world.We outline the structure of FMH anddemonstrate its potential for decentralised learning and control.We find that,given an adequate set of subgoals from which to choose, FMH performs, andparticularly scales, substantially better than cooperative approaches that usea shared reward function.机器学习论文:用于合作强化学习的封建多智能体层次结构(Feudal Multi-Agent Hierarchies for Cooperative Reinforcement Learning) m1140q0k3NN6k50a.jpg
URL地址:https://arxiv.org/abs/1901.08492     ----pdf下载地址:https://arxiv.org/pdf/1901.08492    ----机器学习论文:用于合作强化学习的封建多智能体层次结构(Feudal Multi-Agent Hierarchies for Cooperative Reinforcement Learning)
回复

使用道具 举报

您需要登录后才可以回帖 登录 | 立即注册

本版积分规则 返回列表 发新帖

1449231467当前离线
新手上路

查看:125 | 回复:0

快速回复 返回顶部 返回列表