PHD PROJECTS

Past PHD Projects

**Dissertation:**Computation of the Exact and Approximate Radicals of Ideals: Techniques Based on Matrices of Traces, Moment Matrices and Bezoutians (Itnuit Janovitz Freireich, Ph.D. 2008).

http://repository.lib.ncsu.edu/ir/handle/1840.16/5398

**Description of Project:** The main focus of this project is to solve polynomial systems which have roots with multiplicities or roots forming small clusters. The motivation for this investigation comes from the fact that such systems are considered ill-conditioned and traditional numerical methods are generally unable to handle them. In this project we gave algorithms to compute the radical or approximate radical of a given zero-dimensional ideal based on linear algebra only on a so called matrix of traces.Then we analyzed how these algorithms behave under perturbation of the given ideal or its underlying roots. Finally we gave simple and efficient methods to compute matrices of traces. The results of this project already appeared in papers including: Approximate radical for clusters: a global approach using Gaussian elimination or SVD., Mathematics in Computer Science, Vol. 1, no. 2, pp 393–425, (2007) and Moment matrices, trace matrices and the radical of ideals, Proceedings of the ISSAC 2008, 125--132, ACM, New York, (2008)

**Dissertation:**Tannakian Categories and Linear Differential Algebraic Groups (Alexey Ovchinnikov, Ph.D. 2007)

http://repository.lib.ncsu.edu/ir/handle/1840.16/4801

**Description of Project:** The usual Galois theory of algebraic equations associates a finite group to each polynomial. Properties of this polynomial (e.g., solvability in terms of radicals) are reflected by properties of the groups (e.g., solvability of the group). A similar Galois theory was developed for linear differential equations where the associated groups are linear algebraic groups, that is, groups of matrices whose entries satisfy polynomial equations. Again, properties of the linear differential equation are reflected in properties of the group. Using this Galois theory, algorithms have been developed to determine if a given linear differential equation can be solved in terms of iterations of algebraic functions, exponentials and integrals. The key to many of these algorithms has been an understanding of the representation theory of linear algebraic groups, that is, understanding all the different ways in which a group can be represented as a group of matrices. In the last century, PIerre Deligne gave a category theory based approach to this Galois theory and the representations of these groups. This new way of looking at this subject has been crucial to the algorithms.

In the last 8 years, a Galois theory has been developed for parameterized differential and difference equations. The Galois groups are linear differential algebraic groups, that is, groups of matrices whose entries are functions of the parameters that satisfy certain differential equations. This dissertaion gives a category theoretical approach to deriving this Galois theory and gives a categorical description of the representation theory of these groups with the goal of deriving algorithms in this theory. The results of the thesis have already appeared in print as: Tannakian Categories, Linear Differential Algebraic Groups, and Parametrized Linear Differential Equations, Transformation Groups 14 (1), 2009, 195--223 and Differential Tannakian categories, Journal of Algebra 321 (10), 2009, 3043--3062

**Dissertation:**"Object-Image Correspondence of Under Projections." by Joseph Burdis Ph.D 2010

http://repository.lib.ncsu.edu/ir/handle/1840.16/6109

**Description of Project:**This project is devoted to the problem of establishing an object-image correspondence under parallel and central projections from a 3-dimensional space to a plane. The goal is to develop algorithms for deciding whether there exists a projection that maps a given spatial object to a given object in the plane. The motivation comes from an important problem in computer vision – determining the correspondence between an object and its image obtained by a camera with unknown position, orientation and internal parameters. This dissertation presents new algorithms for deciding projection correspondence for curves and for finite lists of points. Algorithmic solutions to the projection problem for finite lists of points have been previously known and we provide a comparison of the new algorithms, presented in this thesis, with previous methods. An algorithmic solution for the projection problem for curves under parallel or central projections with a large number of unknown parameters appears to be previously unknown.

Department of Mathematics

2108 SAS Hall

2311 Stinson Drive

Box 8205, NC State University

Raleigh, NC 27695-8205

**Phone** (919) 515-2382**
Fax **(919) 513-7336

student resources :

courses

suggested curriculum

phd projects

application procedure

financial assistance

employment resources

courses

suggested curriculum

phd projects

application procedure

financial assistance

employment resources