Files in this item



application/pdfLI-DISSERTATION-2021.pdf (6MB)Restricted to U of Illinois
(no description provided)PDF


Title:Machine learning methodologies for high dimensional biomedical & bioinformatics applications
Author(s):Li, Yutong
Director of Research:Zhu, Ruoqing
Doctoral Committee Chair(s):Zhu, Ruoqing
Doctoral Committee Member(s):Qu, Annie; Zhao, Dave Sihai; Li, Bo
Department / Program:Statistics
Degree Granting Institution:University of Illinois at Urbana-Champaign
Subject(s):Machine Learning
Matrix Decomposition
Deep Learning
Biomedical Applications
Abstract:The impact of machine learning has been greatly expanded due to the increase in computational power in recent years, and has made a significant scientific contribution to many fields. This dissertation primarily investigates and expands the usage of certain machine learning methodologies on high-dimensional biomedical and bioinformatics applications. In particular, I aim to propose novel, data-driven clustering and feature extraction methods to uncover richer and more interpretable predictive features for classification problems. This dissertation considers three modern biomedical and bioinformatics problems in the context of text mining, computer vision and microbiome analysis. To address the different challenges in these applications, novels methods in matrix factorization, image registration, and deep learning are proposed. For the first project on text mining, we propose the semi-orthogonal non-negative matrix factorization as a topic model to investigate the potential of using triage notes to classify patient disposition in addressing the issue of emergency department crowding. For the second project on computer vision, we propose a novel implementation of the neural style transfer algorithm as an image preprocessing and registration method for skin lesion classification problems. For the third project on microbiome analysis, we discuss two works that have been done. First, we propose an analysis pipeline that implements the random forest model to identify food intake, along with a PCA-based approach to remove study batch effects and validate our classification results. Second, we proposed to incorporate the phylogenetic information of microbes as graphs via a graphical convolutional neural network to improve the classification performances for dietary and health outcomes.
Issue Date:2021-04-21
Rights Information:Copyright 2021 by Yutong Li. All rights reserved.
Date Available in IDEALS:2021-09-17
Date Deposited:2021-05

This item appears in the following Collection(s)

Item Statistics