Achintya Kundu, L. Wynter, Rhui Dih Lee, L. A. Bathen
{"title":"transfer -一劳永逸:Edge的AI模型优化","authors":"Achintya Kundu, L. Wynter, Rhui Dih Lee, L. A. Bathen","doi":"10.1109/EDGE60047.2023.00017","DOIUrl":null,"url":null,"abstract":"Weight-sharing neural architecture search aims to optimize a configurable neural network model (supernet) for a variety of deployment scenarios across many devices with different resource constraints. Existing approaches use evolutionary search to extract models of different sizes from a supernet trained on a very large data set, and then fine-tune the extracted models on the typically small, real-world data set of interest. The computational cost of training thus grows linearly with the number of different model deployment scenarios. Hence, we propose Transfer-Once-For-All (TOFA) for supernet-style training on small data sets with constant computational training cost over any number of edge deployment scenarios. Given a task, TOFA obtains custom neural networks, both the topology and the weights, optimized for any number of edge deployment scenarios. To overcome the challenges arising from small data, TOFA utilizes a unified semi-supervised training loss to simultaneously train all subnets within the supernet, coupled with on-the-fly architecture selection at deployment time.","PeriodicalId":369407,"journal":{"name":"2023 IEEE International Conference on Edge Computing and Communications (EDGE)","volume":"20 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2023-03-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Transfer-Once-For-All: AI Model Optimization for Edge\",\"authors\":\"Achintya Kundu, L. Wynter, Rhui Dih Lee, L. A. Bathen\",\"doi\":\"10.1109/EDGE60047.2023.00017\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Weight-sharing neural architecture search aims to optimize a configurable neural network model (supernet) for a variety of deployment scenarios across many devices with different resource constraints. Existing approaches use evolutionary search to extract models of different sizes from a supernet trained on a very large data set, and then fine-tune the extracted models on the typically small, real-world data set of interest. The computational cost of training thus grows linearly with the number of different model deployment scenarios. Hence, we propose Transfer-Once-For-All (TOFA) for supernet-style training on small data sets with constant computational training cost over any number of edge deployment scenarios. Given a task, TOFA obtains custom neural networks, both the topology and the weights, optimized for any number of edge deployment scenarios. To overcome the challenges arising from small data, TOFA utilizes a unified semi-supervised training loss to simultaneously train all subnets within the supernet, coupled with on-the-fly architecture selection at deployment time.\",\"PeriodicalId\":369407,\"journal\":{\"name\":\"2023 IEEE International Conference on Edge Computing and Communications (EDGE)\",\"volume\":\"20 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2023-03-27\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2023 IEEE International Conference on Edge Computing and Communications (EDGE)\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/EDGE60047.2023.00017\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2023 IEEE International Conference on Edge Computing and Communications (EDGE)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/EDGE60047.2023.00017","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Transfer-Once-For-All: AI Model Optimization for Edge
Weight-sharing neural architecture search aims to optimize a configurable neural network model (supernet) for a variety of deployment scenarios across many devices with different resource constraints. Existing approaches use evolutionary search to extract models of different sizes from a supernet trained on a very large data set, and then fine-tune the extracted models on the typically small, real-world data set of interest. The computational cost of training thus grows linearly with the number of different model deployment scenarios. Hence, we propose Transfer-Once-For-All (TOFA) for supernet-style training on small data sets with constant computational training cost over any number of edge deployment scenarios. Given a task, TOFA obtains custom neural networks, both the topology and the weights, optimized for any number of edge deployment scenarios. To overcome the challenges arising from small data, TOFA utilizes a unified semi-supervised training loss to simultaneously train all subnets within the supernet, coupled with on-the-fly architecture selection at deployment time.