Jianqing Cai, Yunfeng Liang, Alexander Knieps, dongkai qi, Erhui Wang, Haoming Xiang, Liang Liao, Jie Huang, Jie Yang, Jia Huang, Jianwen Liu, P. Drews, Shuai Xu, xiang gu, Yichen Gao, Yu Luo, zhi li
{"title":"改进的中断预测神经网络模型训练框架及其在 EXL-50 上的应用","authors":"Jianqing Cai, Yunfeng Liang, Alexander Knieps, dongkai qi, Erhui Wang, Haoming Xiang, Liang Liao, Jie Huang, Jie Yang, Jia Huang, Jianwen Liu, P. Drews, Shuai Xu, xiang gu, Yichen Gao, Yu Luo, zhi li","doi":"10.1088/2058-6272/ad1571","DOIUrl":null,"url":null,"abstract":"\n A neural network model with classical annotation method has been used on EXL-50 tokamak to predict the impending disruptions. However, the results revealed issues of overfitting and overconfidence in predictions caused by the inaccurate labeling. To mitigate these issues, an improved training framework has been proposed. In this approach, soft labels from previous training serve as teachers to supervise the further learning process, which has demonstrated its significant improvement in predictive model performance. Notably, this enhancement is primarily attributed to the coupling effect of the soft labels and correction mechanism. This improved training framework introduces an instance-specific label smoothing method, which reflects a more nuanced model’s assessment on the likelihood of a disruption. It presents a possible solution to effectively address the challenges associated with accurate labeling across different machines","PeriodicalId":20250,"journal":{"name":"Plasma Science & Technology","volume":null,"pages":null},"PeriodicalIF":1.6000,"publicationDate":"2023-12-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"An improved training framework in neural network model fordisruption prediction and its application on EXL-50\",\"authors\":\"Jianqing Cai, Yunfeng Liang, Alexander Knieps, dongkai qi, Erhui Wang, Haoming Xiang, Liang Liao, Jie Huang, Jie Yang, Jia Huang, Jianwen Liu, P. Drews, Shuai Xu, xiang gu, Yichen Gao, Yu Luo, zhi li\",\"doi\":\"10.1088/2058-6272/ad1571\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"\\n A neural network model with classical annotation method has been used on EXL-50 tokamak to predict the impending disruptions. However, the results revealed issues of overfitting and overconfidence in predictions caused by the inaccurate labeling. To mitigate these issues, an improved training framework has been proposed. In this approach, soft labels from previous training serve as teachers to supervise the further learning process, which has demonstrated its significant improvement in predictive model performance. Notably, this enhancement is primarily attributed to the coupling effect of the soft labels and correction mechanism. This improved training framework introduces an instance-specific label smoothing method, which reflects a more nuanced model’s assessment on the likelihood of a disruption. It presents a possible solution to effectively address the challenges associated with accurate labeling across different machines\",\"PeriodicalId\":20250,\"journal\":{\"name\":\"Plasma Science & Technology\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":1.6000,\"publicationDate\":\"2023-12-13\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Plasma Science & Technology\",\"FirstCategoryId\":\"1089\",\"ListUrlMain\":\"https://doi.org/10.1088/2058-6272/ad1571\",\"RegionNum\":3,\"RegionCategory\":\"物理与天体物理\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q3\",\"JCRName\":\"PHYSICS, FLUIDS & PLASMAS\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Plasma Science & Technology","FirstCategoryId":"1089","ListUrlMain":"https://doi.org/10.1088/2058-6272/ad1571","RegionNum":3,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"PHYSICS, FLUIDS & PLASMAS","Score":null,"Total":0}
An improved training framework in neural network model fordisruption prediction and its application on EXL-50
A neural network model with classical annotation method has been used on EXL-50 tokamak to predict the impending disruptions. However, the results revealed issues of overfitting and overconfidence in predictions caused by the inaccurate labeling. To mitigate these issues, an improved training framework has been proposed. In this approach, soft labels from previous training serve as teachers to supervise the further learning process, which has demonstrated its significant improvement in predictive model performance. Notably, this enhancement is primarily attributed to the coupling effect of the soft labels and correction mechanism. This improved training framework introduces an instance-specific label smoothing method, which reflects a more nuanced model’s assessment on the likelihood of a disruption. It presents a possible solution to effectively address the challenges associated with accurate labeling across different machines
期刊介绍:
PST assists in advancing plasma science and technology by reporting important, novel, helpful and thought-provoking progress in this strongly multidisciplinary and interdisciplinary field, in a timely manner.
A Publication of the Institute of Plasma Physics, Chinese Academy of Sciences and the Chinese Society of Theoretical and Applied Mechanics.