Author:  Zhang, Xinzhen 
Title:  Polynomial optimization and its applications 
Degree:  Ph.D. 
Year:  2010 
Subject:  Hong Kong Polytechnic University  Dissertations Polynomials Mathematical optimization 
Department:  Dept. of Applied Mathematics 
Pages:  x, 88 leaves : ill. ; 30 cm. 
InnoPac Record:  http://library.polyu.edu.hk/record=b2393023 
URI:  http://theses.lib.polyu.edu.hk/handle/200/5903 
Abstract:  The main purposes of this thesis are to solve some polynomial optimization problems and to find their applications. The polynomial optimization problems involved in this thesis include the cubic spherical optimization problems and biquadratic optimization problems. The main contributions of this thesis are as follows: In this thesis, we first consider a new model, the truncated generalized diffusion tensor imagine (GDTI) model in medical engineering, which overcomes the drawback that water movement in biological tissues often shows nonGaussian diffusion behavior. In the GDTI model, polynomial associated with even order tensors reflects the magnitude of the signal, while polynomial associated with odd order tensors reflects the phase of the signal. Moreover, we use the apparent skewness coefficient (ASC) value to measure the phase of nonGaussian signals. We present some properties of related tensors and propose a direct computation method for calculating the ASC value. We discuss the general cubic spherical optimization problems, which include the cubic onespherical/twospherical/threespherical optimization problems. For these three problems, we present their NPhardnesses and discuss the complexity results of some special cases. For the NPhardness cases, some approximation solution methods for them are established. Then we study the biquadratic optimization problem over two unit spheres. At first, the problem is equivalently transformed into computing the largest Meigenvalue of related tensor. Based on the reformulation, power method for computing the largest eigenvalue of a matrix is modified to compute the largest Meigenvalue of a tensor. To get a good approximation of the largest Meigenvalue of a tensor, we introduce an initialization technique. The given numerical experiments show that the modified method performs well. Finally, we discuss the biquadratic optimization problems with quadratic constraints. First, we generalize the SDP relaxation scheme for approximately solving NPhard quadratic optimization to solve biquadratic optimization problems. Then we show that each rbound approximation solution of the relaxed bilinear SDP problems can be used to generate in randomized polynomial time an Ϭ(r)approximation solution for biquadratic optimization problems. Furthermore, we show that when the number of constraints is not larger than two, biquadratic optimization problems are equivalent to their corresponding SDP relaxation problems, which generalizes the result in [33]. Then, we present some approximation solutions with some quality bounds for the biquadratic maximization model with some assumptions. For biquadratic optimization problems with two constraints, some approximation solutions are established. Finally, we extend the results from real cases to complex cases. 
Files  Size  Format 

b23930238.pdf  908.7Kb 


As a bona fide Library user, I declare that:  


By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms. 