{"title":"基于多任务特征学习的功能连接图联合特征提取","authors":"A. Altmann, B. Ng","doi":"10.1109/PRNI.2015.17","DOIUrl":null,"url":null,"abstract":"Using sparse regularization in classifier learning is an appealing strategy to locate relevant brain regions and connections between regions within high-dimensional brain imaging data. A major drawback of sparse classifier learning is the lack of stability to data perturbations, which leads to different sets of features being selected. Here, we propose to use multi-task feature learning (MFL) to generate sparse and stable classifiers. In classification experiments on functional connectivity estimated from resting state functional magnetic resonance imaging (fMRI), we show that MFL more consistently selects the same connections across bootstrap samples and provides more interpretable models in multiclass settings than standard sparse classifiers, while achieving similar classification performance.","PeriodicalId":380902,"journal":{"name":"2015 International Workshop on Pattern Recognition in NeuroImaging","volume":"48 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2015-06-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"1","resultStr":"{\"title\":\"Joint Feature Extraction from Functional Connectivity Graphs with Multi-task Feature Learning\",\"authors\":\"A. Altmann, B. Ng\",\"doi\":\"10.1109/PRNI.2015.17\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Using sparse regularization in classifier learning is an appealing strategy to locate relevant brain regions and connections between regions within high-dimensional brain imaging data. A major drawback of sparse classifier learning is the lack of stability to data perturbations, which leads to different sets of features being selected. Here, we propose to use multi-task feature learning (MFL) to generate sparse and stable classifiers. In classification experiments on functional connectivity estimated from resting state functional magnetic resonance imaging (fMRI), we show that MFL more consistently selects the same connections across bootstrap samples and provides more interpretable models in multiclass settings than standard sparse classifiers, while achieving similar classification performance.\",\"PeriodicalId\":380902,\"journal\":{\"name\":\"2015 International Workshop on Pattern Recognition in NeuroImaging\",\"volume\":\"48 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2015-06-10\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"1\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2015 International Workshop on Pattern Recognition in NeuroImaging\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/PRNI.2015.17\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2015 International Workshop on Pattern Recognition in NeuroImaging","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/PRNI.2015.17","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Joint Feature Extraction from Functional Connectivity Graphs with Multi-task Feature Learning
Using sparse regularization in classifier learning is an appealing strategy to locate relevant brain regions and connections between regions within high-dimensional brain imaging data. A major drawback of sparse classifier learning is the lack of stability to data perturbations, which leads to different sets of features being selected. Here, we propose to use multi-task feature learning (MFL) to generate sparse and stable classifiers. In classification experiments on functional connectivity estimated from resting state functional magnetic resonance imaging (fMRI), we show that MFL more consistently selects the same connections across bootstrap samples and provides more interpretable models in multiclass settings than standard sparse classifiers, while achieving similar classification performance.