{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,6,18]],"date-time":"2026-06-18T06:09:06Z","timestamp":1781762946284,"version":"3.54.5"},"reference-count":52,"publisher":"Oxford University Press (OUP)","issue":"5","license":[{"start":{"date-parts":[[2018,9,16]],"date-time":"2018-09-16T00:00:00Z","timestamp":1537056000000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/academic.oup.com\/journals\/pages\/open_access\/funder_policies\/chorus\/standard_publication_model"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2018,11,1]]},"abstract":"<jats:title>Summary<\/jats:title><jats:p>The sparse generalized eigenvalue problem (GEP) plays a pivotal role in a large family of high dimensional statistical models, including sparse Fisher\u2019s discriminant analysis, canonical correlation analysis and sufficient dimension reduction. The sparse GEP involves solving a non-convex optimization problem. Most existing methods and theory in the context of specific statistical models that are special cases of the sparse GEP require restrictive structural assumptions on the input matrices. We propose a two-stage computational framework to solve the sparse GEP. At the first stage, we solve a convex relaxation of the sparse GEP. Taking the solution as an initial value, we then exploit a non-convex optimization perspective and propose the truncated Rayleigh flow method (which we call \u2018rifle\u2019) to estimate the leading generalized eigenvector. We show that rifle converges linearly to a solution with the optimal statistical rate of convergence. Theoretically, our method significantly improves on the existing literature by eliminating structural assumptions on the input matrices. To achieve this, our analysis involves two key ingredients: a new analysis of the gradient-based method on non-convex objective functions, and a fine-grained characterization of the evolution of sparsity patterns along the solution path. Thorough numerical studies are provided to validate the theoretical results.<\/jats:p>","DOI":"10.1111\/rssb.12291","type":"journal-article","created":{"date-parts":[[2018,9,17]],"date-time":"2018-09-17T03:42:41Z","timestamp":1537155761000},"page":"1057-1086","source":"Crossref","is-referenced-by-count":38,"title":["Sparse Generalized Eigenvalue Problem: Optimal Statistical Rates via Truncated Rayleigh Flow"],"prefix":"10.1093","volume":"80","author":[{"given":"Kean Ming","family":"Tan","sequence":"first","affiliation":[{"name":"University of Minnesota , Minneapolis , USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Zhaoran","family":"Wang","sequence":"additional","affiliation":[{"name":"Northwestern University , Evanston , USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Han","family":"Liu","sequence":"additional","affiliation":[{"name":"Northwestern University , Evanston , USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Tong","family":"Zhang","sequence":"additional","affiliation":[{"name":"Tencent Technology Shenzhen \u00a0 People\u2019s Republic of China"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"286","published-online":{"date-parts":[[2018,9,16]]},"reference":[{"key":"2023022002394451300_","first-page":"1269","article-title":"Optimal solutions for sparse principal component analysis","volume":"9","author":"d\u2019Aspremont","year":"2008","journal-title":"J. Mach. Learn. Res."},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"434","DOI":"10.1137\/050645506","article-title":"A direct formulation for sparse PCA using semidefinite programming","volume":"49","author":"d\u2019Aspremont","year":"2007","journal-title":"SIAM Rev"},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"1055","DOI":"10.1214\/12-AOS1014","article-title":"Minimax bounds for sparse PCA with noisy high-dimensional data","volume":"41","author":"Birnbaum","year":"2013","journal-title":"Ann. Statist."},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1561\/2200000016","article-title":"Distributed optimization and statistical learning via the ADMM","volume":"3","author":"Boyd","year":"2010","journal-title":"Foundns Trends Mach. Learn."},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"3074","DOI":"10.1214\/13-AOS1178","article-title":"Sparse PCA: optimal rates and adaptive estimation","volume":"41","author":"Cai","year":"2013","journal-title":"Ann. Statist."},{"key":"2023022002394451300_","volume-title":"Sparse CCA via precision adjusted iterative thresholding","author":"Chen","year":"2013"},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"3696","DOI":"10.1214\/10-AOS826","article-title":"Coordinate-independent sparse sufficient dimension reduction and variable selection","volume":"38","author":"Chen","year":"2010","journal-title":"Ann. Statist."},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"406","DOI":"10.1198\/TECH.2011.08118","article-title":"Sparse discriminant analysis","volume":"53","author":"Clemmensen","year":"2012","journal-title":"Technometrics"},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"2109","DOI":"10.1080\/03610920008832598","article-title":"SAVE: a method for dimension reduction and graphics in regression","volume":"29","author":"Cook","year":"2000","journal-title":"Communs Statist. Theory Meth"},{"key":"2023022002394451300_","first-page":"1","article-title":"Dimension reduction in regression","volume":"22","author":"Cook","year":"2007","journal-title":"Statist. Sci."},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"485","DOI":"10.1214\/08-STS275","article-title":"Principal fitted components for dimension reduction in regression","volume":"23","author":"Cook","year":"2008","journal-title":"Statist. Sci."},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"1187","DOI":"10.1080\/01621459.1999.10473873","article-title":"Dimension reduction in binary response regression","volume":"94","author":"Cook","year":"1999","journal-title":"J. Am. Statist. Ass."},{"key":"2023022002394451300_","volume-title":"Augmented Lagrangian and alternating direction methods for convex optimization: a tutorial and some illustrative computational results","author":"Eckstein","year":"2012"},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"2605","DOI":"10.1214\/07-AOS504","article-title":"High dimensional classification using features annealed independence rules","volume":"36","author":"Fan","year":"2008","journal-title":"Ann. Statist."},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"14","DOI":"10.1214\/14-AOS1307","article-title":"QUADRO: a supervised dimension reduction method via Rayleigh quotient optimization","volume":"43","author":"Fan","year":"2015","journal-title":"Ann. Statist."},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"2168","DOI":"10.1214\/15-AOS1332","article-title":"Minimax estimation in sparse canonical correlation analysis","volume":"43","author":"Gao","year":"2015","journal-title":"Ann. Statist."},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"2074","DOI":"10.1214\/16-AOS1519","article-title":"Sparse CCA: adaptive estimation and computational barriers","volume":"45","author":"Gao","year":"2017","journal-title":"Ann. Statist."},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"2007","DOI":"10.1214\/15-EJS1064","article-title":"Optimal variable selection in multi-group sparse discriminant analysis","volume":"9","author":"Gaynanova","year":"2015","journal-title":"Electron. J. Statist."},{"key":"2023022002394451300_","volume-title":"In Proc. 33rd Int. Conf.","author":"Ge","year":"2016"},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"531","DOI":"10.1126\/science.286.5439.531","article-title":"Molecular classification of cancer: class discovery and class prediction by gene expression monitoring","volume":"286","author":"Golub","year":"1999","journal-title":"Science"},{"key":"2023022002394451300_","volume-title":"Matrix Computations","author":"Golub","year":"2012"},{"key":"2023022002394451300_","author":"Gu","year":"2014"},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"86","DOI":"10.1093\/biostatistics\/kxj035","article-title":"Regularized linear discriminant analysis and its application in microarrays","volume":"8","author":"Guo","year":"2007","journal-title":"Biostatistics"},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"1063","DOI":"10.1109\/TIT.2014.2381241","article-title":"Optimal feature selection in high-dimensional discriminant analysis","volume":"61","author":"Kolar","year":"2015","journal-title":"IEEE Trans. Inform. Theory"},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"417","DOI":"10.1016\/j.compbiolchem.2008.07.015","article-title":"Sparse optimal scoring for multiclass cancer diagnosis and biomarker detection using microarray data","volume":"32","author":"Leng","year":"2008","journal-title":"Computnl Biol. Chem."},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"316","DOI":"10.1080\/01621459.1991.10475035","article-title":"Sliced inverse regression for dimension reduction","volume":"86","author":"Li","year":"1991","journal-title":"J. Am. Statist. Ass."},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"603","DOI":"10.1093\/biomet\/asm044","article-title":"Sparse sufficient dimension reduction","volume":"94","author":"Li","year":"2007","journal-title":"Biometrika"},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"503","DOI":"10.1198\/004017006000000129","article-title":"Sparse sliced inverse regression","volume":"48","author":"Li","year":"2006","journal-title":"Technometrics"},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"124","DOI":"10.1111\/j.1541-0420.2007.00836.x","article-title":"Sliced inverse regression with regularizations","volume":"64","author":"Li","year":"2008","journal-title":"Biometrics"},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"772","DOI":"10.1214\/13-AOS1097","article-title":"Sparse principal component analysis and iterative thresholding","volume":"41","author":"Ma","year":"2013","journal-title":"Ann. Statist."},{"key":"2023022002394451300_","volume-title":"Subspace perspective on canonical correlation analysis: dimension reduction and minimax rates","author":"Ma","year":"2016"},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"134","DOI":"10.1111\/j.1751-5823.2012.00182.x","article-title":"A review on dimension reduction","volume":"81","author":"Ma","year":"2013","journal-title":"Int. Statist. Rev."},{"key":"2023022002394451300_","article-title":"Multiclass sparse discriminant analysis","volume-title":"Statist. Sin.","author":"Mai","year":"2015"},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"29","DOI":"10.1093\/biomet\/asr066","article-title":"A direct approach to sparse discriminant analysis in ultra-high dimensions","volume":"99","author":"Mai","year":"2012","journal-title":"Biometrika"},{"key":"2023022002394451300_","author":"Moghaddam","year":"2006"},{"key":"2023022002394451300_","author":"Moghaddam","year":"2006"},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"1005","DOI":"10.1080\/10618600.2015.1073155","article-title":"Fused lasso additive model","volume":"25","author":"Petersen","year":"2016","journal-title":"J. Computnl Graph. Statist."},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"361","DOI":"10.1038\/nm1556","article-title":"Airway epithelial gene expression in the diagnostic evaluation of smokers with suspect lung cancer","volume":"13","author":"Spira","year":"2007","journal-title":"Nat. Med."},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"69","DOI":"10.1016\/0024-3795(79)90094-6","article-title":"Pertubation bounds for the definite generalized eigenvalue problem","volume":"23","author":"Stewart","year":"1979","journal-title":"Lin. Alg. Appl."},{"key":"2023022002394451300_","volume-title":"Matrix Perturbation Theory","author":"Stewart","year":"1990"},{"key":"2023022002394451300_","doi-asserted-by":"crossref","DOI":"10.1093\/biomet\/asy049","article-title":"A convex formulation for high-dimensional sparse sliced inverse regression","author":"Tan","year":"2018","journal-title":"Biometrika"},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"104","DOI":"10.1214\/ss\/1056397488","article-title":"Class prediction by nearest shrunken centroids, with applications to DNA microarrays","volume":"18","author":"Tibshirani","year":"2003","journal-title":"Statist. Sci."},{"key":"2023022002394451300_","author":"Vu","year":"2013"},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"2905","DOI":"10.1214\/13-AOS1151","article-title":"Minimax sparse principal subspace estimation in high dimensions","volume":"41","author":"Vu","year":"2013","journal-title":"Ann. Statist."},{"key":"2023022002394451300_","author":"Wang","year":"2013"},{"key":"2023022002394451300_","author":"Wang","year":"2014"},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"515","DOI":"10.1093\/biostatistics\/kxp008","article-title":"A penalized matrix decomposition, with applications to sparse principal components and canonical correlation analysis","volume":"10","author":"Witten","year":"2009","journal-title":"Biostatistics"},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"879","DOI":"10.1111\/rssb.12093","article-title":"Sequential sufficient dimension reduction for large p, small n problems","volume":"77","author":"Yin","year":"2015","journal-title":"J. R. Statist. Soc."},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"315","DOI":"10.1093\/biomet\/asv008","article-title":"A useful variant of the Davis\u2013Kahan theorem for statisticians","volume":"102","author":"Yu","year":"2014","journal-title":"Biometrika"},{"key":"2023022002394451300_","first-page":"899","article-title":"Truncated power method for sparse eigenvalue problems","volume":"14","author":"Yuan","year":"2013","journal-title":"J. Mach. Learn. Res."},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"630","DOI":"10.1198\/016214505000001285","article-title":"On sliced inverse regression with high-dimensional covariates","volume":"101","author":"Zhu","year":"2006","journal-title":"J. Am. Statist. Ass."},{"key":"2023022002394451300_","doi-asserted-by":"crossref","first-page":"265","DOI":"10.1198\/106186006X113430","article-title":"Sparse principal component analysis","volume":"15","author":"Zou","year":"2006","journal-title":"J. Computnl Graph. Statist."}],"container-title":["Journal of the Royal Statistical Society Series B: Statistical Methodology"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/api.wiley.com\/onlinelibrary\/tdm\/v1\/articles\/10.1111%2Frssb.12291","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/academic.oup.com\/jrsssb\/article-pdf\/80\/5\/1057\/49269636\/jrsssb_80_5_1057.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/academic.oup.com\/jrsssb\/article-pdf\/80\/5\/1057\/49269636\/jrsssb_80_5_1057.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2023,9,5]],"date-time":"2023-09-05T04:39:50Z","timestamp":1693888790000},"score":1,"resource":{"primary":{"URL":"https:\/\/academic.oup.com\/jrsssb\/article\/80\/5\/1057\/7048354"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2018,9,16]]},"references-count":52,"journal-issue":{"issue":"5","published-online":{"date-parts":[[2018,9,16]]},"published-print":{"date-parts":[[2018,11,1]]}},"URL":"https:\/\/doi.org\/10.1111\/rssb.12291","relation":{},"ISSN":["1369-7412","1467-9868"],"issn-type":[{"value":"1369-7412","type":"print"},{"value":"1467-9868","type":"electronic"}],"subject":[],"published-other":{"date-parts":[[2018,11]]},"published":{"date-parts":[[2018,9,16]]}}}