{"title":"一种数据驱动的深度神经网络剪枝方法用于有效的数字信号调制识别","authors":"Ya Tu, Meiyu Wang, Sen Wang","doi":"10.1109/GCWkshps45667.2019.9024610","DOIUrl":null,"url":null,"abstract":"State-of-the-art digital signal modulation recognition approaches involves more and more deep learning method. However, the tremendous potential exhibited by deep learning is often offset by architectural and computational complexity, making widespread deployment a challenge for edge scenarios such as mobile and other consumer devices. To the best of our knowledge, this is almost the first work to deploy deep neural work in edge equipment for physical signal classification. Inspired by an observation that fully connected layer makes up about 75% of total parameters, we will focus on network pruning in deep neural network fully connected layer. Our experiments show that we could obtain high compression ratio about 17%~18% of parameters without losing over 0.1% in accuracy.","PeriodicalId":210825,"journal":{"name":"2019 IEEE Globecom Workshops (GC Wkshps)","volume":null,"pages":null},"PeriodicalIF":0.0000,"publicationDate":"2019-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"2","resultStr":"{\"title\":\"A Data-Driven Deep Neural Network Pruning Approach Towards Efficient Digital Signal Modulation Recognition\",\"authors\":\"Ya Tu, Meiyu Wang, Sen Wang\",\"doi\":\"10.1109/GCWkshps45667.2019.9024610\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"State-of-the-art digital signal modulation recognition approaches involves more and more deep learning method. However, the tremendous potential exhibited by deep learning is often offset by architectural and computational complexity, making widespread deployment a challenge for edge scenarios such as mobile and other consumer devices. To the best of our knowledge, this is almost the first work to deploy deep neural work in edge equipment for physical signal classification. Inspired by an observation that fully connected layer makes up about 75% of total parameters, we will focus on network pruning in deep neural network fully connected layer. Our experiments show that we could obtain high compression ratio about 17%~18% of parameters without losing over 0.1% in accuracy.\",\"PeriodicalId\":210825,\"journal\":{\"name\":\"2019 IEEE Globecom Workshops (GC Wkshps)\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2019-12-01\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"2\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2019 IEEE Globecom Workshops (GC Wkshps)\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/GCWkshps45667.2019.9024610\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2019 IEEE Globecom Workshops (GC Wkshps)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/GCWkshps45667.2019.9024610","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
A Data-Driven Deep Neural Network Pruning Approach Towards Efficient Digital Signal Modulation Recognition
State-of-the-art digital signal modulation recognition approaches involves more and more deep learning method. However, the tremendous potential exhibited by deep learning is often offset by architectural and computational complexity, making widespread deployment a challenge for edge scenarios such as mobile and other consumer devices. To the best of our knowledge, this is almost the first work to deploy deep neural work in edge equipment for physical signal classification. Inspired by an observation that fully connected layer makes up about 75% of total parameters, we will focus on network pruning in deep neural network fully connected layer. Our experiments show that we could obtain high compression ratio about 17%~18% of parameters without losing over 0.1% in accuracy.