SVD Approximations of Large Scale Inverse Problems Pubblico

Meng, Chang (2016)

Permanent URL: https://etd.library.emory.edu/concern/etds/fb494892v?locale=it
Published

Abstract

This thesis studies efficient methods for computing approximations of the singular value decomposition (SVD) for large matrices that arise in ill-posed inverse problems, with a focus on image deblurring. These methods are: the Lanczos method, the randomized method, and the Kronecker product approximation method. After introducing the SVD and describing the approximation methods, we show test results involving the accuracy and speed comparisons of these methods, and provide some deblurring examples.

Table of Contents

1 Introduction. 1

2 Singular Value Decomposition. 7

2.1 Definition and Properties. 7

2.2 Computational Difficulties. 9

2.3 Regularization Methods. 13

2.3.1 Truncated SVD. 13

2.3.2 Tikhonov Regularization. 14

3 SVD Approximations. 19

3.1 Lanczos Method. 19

3.2 Randomized Method. 20

3.3 Kronecker Product Approximation Method. 22

4 Numerical Experiments. 25

4.1 Accuracy Test. 25

4.2 Speed Test. 30

4.3 Deblurring Examples. 33

5 Conclusion and Discussion 40

About this Honors Thesis

Rights statement
  • Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.
School
Department
Degree
Submission
Language
  • English
Research Field
Parola chiave
Committee Chair / Thesis Advisor
Committee Members
Ultima modifica

Primary PDF

Supplemental Files