arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2202.05129cs.AIcs.HC
收藏

Help Me Explore:基于图的自动化代理的最小社会干预

Help Me Explore: Minimal Social Interventions for Graph-Based Autotelic Agents

Sorbonne University(索邦大学);INRIA(法国国家信息与自动化研究所)

Ahmed Akakzia, Olivier Serris, Olivier Sigaud, Cédric Colas

摘要:

在寻求自主智能体学习开放式技能的过程中,大多数作品都采用了皮亚杰的观点:学习轨迹是发育智能体与其物理环境之间相互作用的结果。另一方面,维果茨基的观点强调社会文化环境的中心地位:高级认知功能来自于被主体内化的社会文化过程的传递。本文认为,这两个角度可以耦合在学习的autotelic代理,以促进他们的技能习得。为此,我们做了两个贡献:1)一个新的社会互动协议,称为帮助我探索(HME),其中autotelic代理可以受益于个人和社会引导的探索。在社会事件中,一个社会伙伴提出了学习主体知识前沿的目标。在autotelic情节中,代理可以学习掌握自己发现的目标或自主排练失败的社会目标; 2)GANGSTR,一个基于图形的autotelic代理,能够将目标分解为中间子目标序列的操纵域。我们表明,当在HME内学习时,GANGSTR通过掌握最复杂的配置(例如5块的堆栈),仅需很少的社会干预,克服了其个人学习限制。

英文摘要:

In the quest for autonomous agents learning open-ended repertoires of skills, most works take a Piagetian perspective: learning trajectories are the results of interactions between developmental agents and their physical environment. The Vygotskian perspective, on the other hand, emphasizes the centrality of the socio-cultural environment: higher cognitive functions emerge from transmissions of socio-cultural processes internalized by the agent. This paper argues that both perspectives could be coupled within the learning of autotelic agents to foster their skill acquisition. To this end, we make two contributions: 1) a novel social interaction protocol called Help Me Explore (HME), where autotelic agents can benefit from both individual and socially guided exploration. In social episodes, a social partner suggests goals at the frontier of the learning agent knowledge. In autotelic episodes, agents can either learn to master their own discovered goals or autonomously rehearse failed social goals; 2) GANGSTR, a graph-based autotelic agent for manipulation domains capable of decomposing goals into sequences of intermediate sub-goals. We show that when learning within HME, GANGSTR overcomes its individual learning limits by mastering the most complex configurations (e.g. stacks of 5 blocks) with only few social interventions.

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。

↑