“工程应用”导向的类脑计算实验教学研究——以BP神经网络为例

“Engineering Applications”Oriented Brain-Inspired Computing Experimental Teaching Research:Taking BP Neural Network as an Example

针对“类脑计算与芯片”课程学生对抽象的深度学习理论理解困难且动手能力不足的情况,设计了以工程应用为导向的类脑计算教学实验:以BP神经网络为例,开发了基于深度学习的数码管数字识别实验;并在课堂上采用工程应用型一体化教学模式:在教学中以实际应用为案例,做到理论知识与工程应用结合;在训练单元采取“基础训练+项目驱动”的方式,评价方式采用过程和结果综合考核。通过该实验及教学模式的革新,学生加深了对深度学习原理的理解,熟悉了利用嵌入式平台进行“云端训练+边缘推断”深度学习工程应用的开发流程;提高了学习的主观能动性和工程实践能力。

Aiming at promoting the students’understanding of deep learning and their practical ability,we design“Engineering Applications”oriented Brain-inspired computing experimental course.Taking BP neural network as an example,we design an experiment of deep learning for digital tube recognition.In this teaching mode,teachers take the engineering products as examples,extract questions and highlights,and guide the students to think independently.In this way theoretical knowledge is incorporated to the engineering training.Besides,“basic training+project-driven”approaches are used in the training sections.And the evaluation method adopts a comprehensive assessment method of process and results.It is expected that via this experimental course,not only will the students’understanding of deep learning be enhanced,but also they will be able to develop‘cloud computing+edge inference’style deep learning applications on embedded systems.