学习科学视角下深度学习的多模态研究

The Multi-modal Study of Deep Learning from the Perspective of Learning Science

文章从技术、推广、评价、整合和拓展五个方面,说明深度学习研究有赖于基本规律研究。作为学习科学的重要研究内容,深度学习基本规律研究应借鉴多模态的研究方法,全面揭示深度学习发生的内在机制。同时,文章归纳了不同学科和不同行为的多模态,着重分析了学习科学与深度学习领域的多模态。在完善深度学习框架的基础上,文章针对框架的各个组成部分提出了基础型、专业型多模态指标,构建了基于深度学习框架的多模态指标体系,可为不同目的、条件和层次的深度学习基本规律的多模态研究提供指标依据和机制参考。

This paper proposed that the study of deep learning depended on the research of the basic laws from the five aspects of technology, promotion, assessment, integration and expansion. As an important part of the learning science, the research of the basic laws of deep learning should draw on the multi-modal research method, in order to fully reveal the internal mechanism of deep learning. Meanwhile, the multi-mode of different disciplines and behaviors were summarized, in which the multi-mode of learning science and deep learning were emphatically analyzed. The basic and professional multi-modal indexes for various components of the framework were also proposed on the basis of perfecting the framework of deep learning. Moreover, the multi-modal index system based on the framework of deep learning was constructed, which can provide index basis and mechanism reference for the multi-modal study of the basic laws of deep learning with different purposes, conditions and levels.