项目名称: 张量最优化算法及其在基因表达数据中的应用
项目编号: No.11301436
项目类型: 青年科学基金项目
立项/批准年度: 2014
项目学科: 数理科学和化学
项目作者: 陈碧连
作者单位: 厦门大学
项目金额: 22万元
中文摘要: 以张量为工具来分析和研究现实中的问题越来越受到关注,其应用领域非常广泛,包括计算机视觉、数据挖掘、信号处理、神经系统科学、图形分析,生物医学工程、量子力学等等,张量分解与逼近问题是重要的分析依据之一。张量计算是应用数学中一个新的分支,它与多项式优化、张量优化等问题紧密相联。本项目研究的重点在于构造张量分解与逼近的优化模型、揭示张量复杂结构下蕴含的性质、应用于生物信息学中基因表达数据的分析,其中所涉及的优化问题都为NP困难的。目前国内外研究在求解张量分解与逼近问题中具备收敛性且高效性的算法尚不多见,在新兴学科生物信息学方面的实际应用缺乏。本项目立志于给出操作性强、运行时间快且有收敛性保证的算法,设计新的张量优化模型以解决实际中的问题,搭建张量优化与多项式优化之间的桥梁,这一研究既有理论上的深刻性又有应用前景的广泛性,同时充实相关问题的理论与算法,促进相关学科的发展。
中文关键词: 张量优化;分块坐标下降算法;低秩逼近;多项式优化;基因数据表达
英文摘要: Much research attention has been attracted by using tensor as a tool to analyze and study real problems. It has applications in a broad range of areas, including computer vision, data mining, signal processing, neuroscience, graphical analysis, biomedical engineering, quantum mechanics, etc. Tensor decomposition and approximation is one of the important analyzing methods there. Tensor calculus is a new branch of applied mathematics, which has close relationship with polynomial optimization and tensor optimization problems. The project focuses on establishing the optimization models of tensor decomposition and approximation problems, revealing the inherent nature of complex structured tensors, and analyzing gene expression data in bioinformatics, while the optimization problems involved are all NP-hard in general. Up till now, the algorithms with convergence and efficiency for solving tensor decomposition and approximation problems are rare, and the practical applications found in the area of bioinformatics are still lack. Based on these, this project aspires to provide some easy-to-implement and efficient algorithms with convergence guaranteed, design new tensor optimization models for solving real problems, and build the bridge between tensor optimization and polynomial optimization problems, which will bring b
英文关键词: Tensor optimization;Block coordinate descent algorithm;Low-rank approximation;Polynomial optimization;Gene expression data