M. Abrishami, Amir Erfan Eshratifar, D. Eigen, Yanzhi Wang, Shahin Nazarian, M. Pedram
{"title":"基于嵌入空间增强的深度卷积神经网络高效训练","authors":"M. Abrishami, Amir Erfan Eshratifar, D. Eigen, Yanzhi Wang, Shahin Nazarian, M. Pedram","doi":"10.1109/ISQED48828.2020.9137053","DOIUrl":null,"url":null,"abstract":"Recent advances in the field of artificial intelligence have been made possible by deep neural networks. In applications where data are scarce, transfer learning and data augmentation techniques are commonly used to improve the generalization of deep learning models. However, fine-tuning a transfer model with data augmentation in the raw input space has a high computational cost to run the full network for every augmented input. This is particularly critical when large models are implemented on embedded devices with limited computational and energy resources. In this work, we propose a method that replaces the augmentation in the raw input space with an approximate one that acts purely in the embedding space. Our experimental results show that the proposed method drastically reduces the computation, while the accuracy of models is negligibly compromised.","PeriodicalId":225828,"journal":{"name":"2020 21st International Symposium on Quality Electronic Design (ISQED)","volume":"119 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2020-02-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"1","resultStr":"{\"title\":\"Efficient Training of Deep Convolutional Neural Networks by Augmentation in Embedding Space\",\"authors\":\"M. Abrishami, Amir Erfan Eshratifar, D. Eigen, Yanzhi Wang, Shahin Nazarian, M. Pedram\",\"doi\":\"10.1109/ISQED48828.2020.9137053\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Recent advances in the field of artificial intelligence have been made possible by deep neural networks. In applications where data are scarce, transfer learning and data augmentation techniques are commonly used to improve the generalization of deep learning models. However, fine-tuning a transfer model with data augmentation in the raw input space has a high computational cost to run the full network for every augmented input. This is particularly critical when large models are implemented on embedded devices with limited computational and energy resources. In this work, we propose a method that replaces the augmentation in the raw input space with an approximate one that acts purely in the embedding space. Our experimental results show that the proposed method drastically reduces the computation, while the accuracy of models is negligibly compromised.\",\"PeriodicalId\":225828,\"journal\":{\"name\":\"2020 21st International Symposium on Quality Electronic Design (ISQED)\",\"volume\":\"119 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2020-02-12\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"1\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2020 21st International Symposium on Quality Electronic Design (ISQED)\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/ISQED48828.2020.9137053\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2020 21st International Symposium on Quality Electronic Design (ISQED)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ISQED48828.2020.9137053","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Efficient Training of Deep Convolutional Neural Networks by Augmentation in Embedding Space
Recent advances in the field of artificial intelligence have been made possible by deep neural networks. In applications where data are scarce, transfer learning and data augmentation techniques are commonly used to improve the generalization of deep learning models. However, fine-tuning a transfer model with data augmentation in the raw input space has a high computational cost to run the full network for every augmented input. This is particularly critical when large models are implemented on embedded devices with limited computational and energy resources. In this work, we propose a method that replaces the augmentation in the raw input space with an approximate one that acts purely in the embedding space. Our experimental results show that the proposed method drastically reduces the computation, while the accuracy of models is negligibly compromised.