Sparse Signal Recovery with Subbotin Noise

It appears your Web browser is not configured to display PDF files. Download adobe Acrobat or click here to download the PDF file.

Click here to download the PDF file.

Creator: 

Miller, Joshua Charles

Date: 

2022

Abstract: 

In this thesis, we expand upon the work of Cui [4] and earlier authors by studying the problem of variable selection in a high-dimensional setting. Specifically, we consider a single vector X from a sequence model with noise components originating from a generalized normal distribution and determine the regions of exact and almost full selection with respect to a Hamming loss function. We make the routine assumption that the model studied is sparse. That is, the informative number of components is tiny in the model relative to the dimension d. An adaptive procedure is also proposed for estimating the signal components of X when the level of sparsity is unknown. A synthetic simulation and empirical study is then presented to showcase the aforementioned results. Lastly, we conclude the thesis by proposing areas of future work.

Subject: 

Statistics
Estimation theory

Language: 

English

Publisher: 

Carleton University

Thesis Degree Name: 

Master of Science: 
M.Sc.

Thesis Degree Level: 

Master's

Thesis Degree Discipline: 

Statistics

Parent Collection: 

Theses and Dissertations

Items in CURVE are protected by copyright, with all rights reserved, unless otherwise indicated. They are made available with permission from the author(s).