Tensor Based Dinkelbach Method for Homogeneous Single Ratio Fractional Programming

报告摘要：In this report, we consider a kind of homogeneous single ration fractional programming. We prove that the crucial subproblem of Dinkelbach-type method is equivalent with a multilinear optimization problem. Then an efficient proximal alternative minimization method is introduced to solve the homogeneous single ration fractional programming with spherical constraint. The subsequential and global sequence convergence of the algorithm are established. Furthermore, we reformulate the equivalent polynomial programming to an equivalent unconstraint programming and the convergence rate of the proposed algorithm is given based on KL inequality. Finally, the proposed algorithm is applied to compute the generalized eigenvalue relative to two symmetric tensors.

报告人简介：陈海滨，教授，博导，香港理工大学博士。国家自然科学基金函评专家、教育部长江学者函评专家，山东省杰青、山东省青年创新团队“张量大数据分析与优化”负责人、山东省取得突出成绩的博士后。兼任中国运筹学会数学规划分会、算法软件及应用分会理事。先后主持国家自然科学基金2项、省部级项目6项，发表SCI论文50多篇.先后多次受邀到香港理工大学，香港城市大学，香港智能多维数据分析中心，澳大利亚科廷大学访学交流。