Deep Reinforcement Learning of Abstract Reasoning from Demonstrations
Deep Reinforcement Learning of Abstract Reasoning from Demonstrations
复制标题
从演示中进行抽象推理的深度强化学习
DOI：
10.1145/3171221.3171289
复制
发表时间：
2018
期刊：
影响因子：
--
通讯作者：
M. Begum
中科院分区：
文献类型：
--
作者：
Madison Clark;M. Begum
仅仅根据观察就能提取出一套可概括的规则，来管理人类之间复杂的、高层次的互动，这是一种高层次的认知能力。掌握这种技能标志着人类发展过程中的一个重要里程碑。在自主机器人中设计这种能力的一个关键挑战是发现歧视性特征之间的关系。识别自然场景中代表特定事件或交互的特征（即，“区别性特征”），然后发现关系（例如，时间/空间/时空/因果）是非平凡的问题。它们往往表现为“鸡生蛋还是蛋生鸡”的困境。本文提出了一个端到端的学习框架，以解决这两个问题的背景下，学习广义的，高层次的规则，人类互动的结构化演示。我们采用我们提出的深度强化学习框架来学习一组规则，这些规则基于对会话的几个实例的观察来管理两个代理之间的行为干预会话。我们还测试了我们的框架与人类受试者在不同情况下的准确性。
Extracting a set of generalizable rules that govern the dynamics of complex, high-level interactions between humans based only on observations is a high-level cognitive ability. Mastery of this skill marks a significant milestone in the human developmental process. A key challenge in designing such an ability in autonomous robots is discovering the relationships among discriminatory features. Identifying features in natural scenes that are representative of a particular event or interaction (i.e. ‘discriminatory features’) and then discovering the relationships (e.g., temporal/spatial/spatiotemporal/causal) among those features in the form of generalized rules are non-trivial problems. They often appear as a ’chicken-and-egg’ dilemma. This paper proposes an end-to-end learning framework to tackle these two problems in the context of learning generalized, high-level rules of human interactions from structured demonstrations. We employed our proposed deep reinforcement learning framework to learn a set of rules that govern a behavioral intervention session between two agents based on observations of several instances of the session. We also tested the accuracy of our framework with human subjects in diverse situations.
做我想做的事，而不是我做过的事：通过规划行动序列来模仿技能
DOI：
10.1109/iros.2016.7759556
发表时间：
2016
期刊：
IROS
影响因子：
--
作者：
Paxton, Chris;Jonathan, Felix;Kobilarov, Marin;Hager, Gregory D.
通讯作者：
Hager, Gregory D.