当前位置: > 学院通知 > 学术报告 >
Structured Sparse SVD Models and Their Applications
时间:2016-11-30 点击:
Structured Sparse SVD Models and Their Applications
for Pattern Discovery
Shihua Zhang
Academy of Mathematics and Systems Science, Chinese Academy of Sciences
Email: zsh@amss.ac.cn
 
Abstract
Learning the ``blocking'' pattern is a central challenge for high dimensional data (e.g., gene expression data). Recently, the Sparse Singular Value Decomposition (SSVD) has been used to achieve this goal. However, this model ignores the structural information between variables (e.g., protein-protein interaction [PPI] network or graph). Although typical graph-regularized norm can incorporate such priori graph information to get accurate discovery and better interpretability, it fails to consider the opposite effect of variables with different signs. Here we propose a novel sparse graph-regularized SVD framework as a powerful tool for high-dimensional data analysis and pattern discovery. The key of this method is to impose a novel penalty which combines a novel graph-regularized norm and L1-norm (or L0-norm) on singular vectors to induce structured sparsity. However, such a non-convex graph-regularized penalty proposes new challenges on model solving. We propose a trick to cleverly remove the absolute operator and design an efficient Alternating Iterative Sparse Projection (AISP) algorithm to solve it. We further propose group sparse SVD models to consider the effect of prior (overlapping) group structured knowledge. We also adopt a partial least square regression method (which is also a SVD related method) for exploring gene-drug co-modules from gene expression and drug response data. The results on synthetic data and real data show that our methods are more effective than other SVD-based methods in terms of specificity and sensitivity.
 
个人简介:

张世华,现任中国科学院数学与系统科学研究院副研究员、中国科学院随机复杂结构与数据科学重点实验室副主任、中国科学院大学岗位教授。主要从事模式识别与生物信息学交叉研究。目前担任BMC Genomics,Scientific Reports,Current Bioinformatics和Frontiers in BCB等杂志的编委以及IEEE/ACM TCBB的客座编辑。曾经荣获中国青年科技奖、中国科学院卢嘉锡青年人才奖、钟家庆运筹学奖、全国百篇优秀博士论文奖、中国科学院院长奖特别奖等;获得国家自然科学基金“优秀青年”基金、中国科学院“卓越青年科学家”项目、入选“陈景润未来之星”特殊人才计划、中国科学院青年创新促进会兼任首届数理分会会长。

报告时间:2016年12月8日9:30-10:30
报告地点:北校区主楼四区107室