Atomistic and Holistic Learning

原子性和整体性学习

• 批准号: 9976739
• 负责人: Robert Cummins
• 金额: $ 12.14万
• 依托单位: University of California-Davis
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 1999
• 资助国家: 美国
• 起止时间: 1999-07-15 至 2001-06-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=9976739&HistoricalAwards=false
• 关键词: Atomistic; Holistic; Learning

Learning is typically characterized as the acquisition of new beliefs, and an assumption that is usually made is that these beliefs can be acquired one at a time. To the extent that we learn in this atomistic fashion, this poses a threat to models of learning that use fully distributed neural nets. However, fully distributed neural nets are effective at learning to perform certain tasks that have been intractable in classical AI, e.g., tasks relating to perception and motor control. The goal of the research proposed here is to investigate the properties of these two types of learning and their relation to each other. We intend to do so by identifying the appropriate task domains of atomistic and holistic learning, by characterizing atomistic and holistic increases or decreases in knowledge, and by working toward a better understanding of whether and how atomistic and holistic learning are related.This research is important both in psychology and the philosophy of science. Many psychological explanations of cognitive processes assign a central role to learning. If we are to eventually understand these explanations in neural terms, we must be able to say not only how, in general, knowledge is acquired, e.g., whether this acquisition is holistic or atomistic (or a combination thereof), but under what conditions this acquisition takes place. Our research proposes to advance current naturalistic understanding of science by detailing the relation between science, which is a paradigmatic example of an atomistic learner, and its cognitive basis, which, if our previously funded research is to be believed, is only partly capable of atomistic learning. We hope this investigation will shed light on some issues in epistemology since we cannot hope to even articulate an epistemology appropriate to neurally based cognition until we have a viable account both of its proprietary forms of learning and of their relation.

学习通常被描述为获得新的信念，并且通常做出的假设是这些信念可以一次获得一个。就我们以这种原子方式学习而言，这对使用完全分布式神经网络的学习模型构成了威胁。然而，完全分布式神经网络可以有效地学习执行经典人工智能中难以处理的某些任务，例如与感知和运动控制相关的任务。这里提出的研究的目的是调查这两种学习类型的属性及其相互关系。我们打算通过确定原子学习和整体学习的适当任务领域、描述知识的原子学习和整体学习的增加或减少、以及努力更好地理解原子学习和整体学习是否以及如何相关来实现这一目标。这项研究很重要在心理学和科学哲学中。许多对认知过程的心理学解释都赋予学习以核心作用。如果我们最终要用神经术语理解这些解释，我们不仅必须能够说明一般情况下如何获取知识，例如这种获取是整体的还是原子的（或其组合），而且能够说明在什么条件下这种获取收购发生。我们的研究旨在通过详细说明科学（原子论学习者的典型例子）与其认知基础之间的关系来推进当前对科学的自然主义理解，如果我们之前资助的研究可信的话，科学基础只能部分地进行原子论学习。学习。我们希望这项研究能够阐明认识论中的一些问题，因为在我们对其专有的学习形式及其关系有一个可行的解释之前，我们甚至不能指望阐明一种适合基于神经的认知的认识论。