报告时间：2015.4.17（星期五）下午14：30
报告地点：博习楼327
报告人：Ivica Kopriva Ph.D
邀请人：陈新建 特聘教授
报告人简介：Ivica Kopriva received PhD degree from the Faculty of the Electrical Engineering and Computing, University of Zagreb in 1998 in the field of signal processing. He has been senior research scientist with the ECE Department, The George Washington University, Washington, DC, USA, 2001-2005. Since 2006, he is a senior scientist at the Ru?er Boškovi? Institute, Zagreb, Croatia. His research is focused on theory and applications of inverse problems, most notably blind source separation, in imaging, spectroscopy and variable selection. He has co-authored around 40 papers in internationally recognized journals and research monograph: Kernel Based Algorithms for Mining Huge Data Sets: Supervised, Semi-supervised and Unsupervised Learning (Springer-Verlag, 2006). He has received 2009 state award for science of the Republic of Croatia, and 2010, 2011 and 2012 awards of the director of Ru?er Boškovi? Institute for publications in high impact factor journals and for competitive grant from Croatian Science Foundation. He has been visiting scientist of the Brain Science Institute, RIKEN, Saitama, Japan in October 2011, and visiting professor of the University of South Toulon du Var, La Garde, France, in April 2012. He holds three US patents and one Canadian patent.
报告摘要：The talk will present most recent development in methods for solving sparseness and nonnegativity constrained nonlinear underdetermined blind source separation (uBSS) problem. The methodology is based on mapping original nonlinear uBSS problem onto reproducible kernel Hilbert space and executing sparseness and nonnegativity constrained linear uBSS problem therein. The method will be demonstrated on numerical examples and on demanding experimental problems related to: unsupervised (automated) segmentation of tissues (resp. organs) from multispectral (resp. CT) images, unsupervised extraction of pure components (analytes) from mass spectra of nonlinear chemical reactions as well as on variable (feature) selection in genomics and proteomics.