摘要
卷积神经网络的轻量化研究工作正成为热门方向,根据其相关理论技术,用人工设计方法和神经网络结构搜索方法对常用的7个轻量化模型进行了剖析,并归纳了它们的核心创新点。最后以此为基础,对该方向的发展作了展望。
The lightweight research work of convolutional neural networks is becoming a popular direction. According to its related theoretical techniques,7 commonly used lightweight models are analyzed from artificial design methods and neural architecture search methods in this paper, and their core innovations point is also summarized. Finally, based on above, the development of this direction is prospected.
出处
《工业控制计算机》
2022年第10期109-111,114,共4页
Industrial Control Computer
基金
湖北省教育厅高校教研项目(2017299)
江汉大学研究生科研创新基金(210852034)。
关键词
卷积神经网络
轻量化
人工设计
神经网络结构搜索
convolutional neural networks
lightweight
artificial design
neural architecture search
