学习架构：深度学习灵活性表达

Learning Architecture: An Expression of Deep Learning Flexibility

针对深度学习的灵活性诉求,研究尝试构建一种灵活的框架形式——学习架构。为此,研究从“架构”的词义以及其在建筑学领域与计算机科学领域的含义入手,经过两次探究分析并类比教学结构的定义提出了表征灵活性的学习架构界定。之后研究论述了三种现有学习架构模型,并从深度学习特征的角度进行解析,得出它们最多只能关注深度学习某方面特征的结论。基于这些认识,研究构建了一种能够表征学习任务、学习活动、学习进程、教学决策四方面灵活性的深度学习架构模型,并论述了其理念机制。希望研究能够加深学者对学习架构的认知,为深度学习的研究提供新方向。

In order to meet the demand for flexibility of deep learning,this study aims to build a flexible framework:learning architecture.In this regard,starting with the meaning of"architecture"in the fields of architecture and computer science,this study proposes the definition of learning architecture representing flexibility based on the exploration and analogy with the definition of pedagogical structure.Then,this study further analyzes three existing learning architecture models from the perspective of deep learning,and finds that they can only meet some features of deep learning.Based on those understandings,this study constructs a deep learning architecture model that can reflect the flexibility of learning tasks,learning activities,learning process and teaching decisions,and discusses its conceptual mechanism.It is hoped that this study can deepen scholars'understanding of the learning architecture and provide a new direction for the research of deep learning.