Matlab Based Machine Learning Software for Education
- Abstract
- Machine learning, machine intelligence and artificial intelligence has attracted tremendous attention from industry and academia alike. The rise of machine learning is powered by none other than the zealous researchers who want to create intelligence machines. The topic has grabbed so much attention that machine learning has becoming a must-to-learn and must-to-acquire knowledge domain of today academia. Matlab being one of the most sought after platform for machine learning has attracted considerable attention from researchers. There are multiple matlab based toolboxes available aimed at teaching machine learning. Many of these toolboxes lack GUI support and coherent implementation. This dissertation is an effort to help young researchers grab some basic concepts of machine learning, especially classification through an easy-to-use matlab GUI platform. The classifiers discussed in this document are divided into two categories; shallow classifiers and deep classifiers. Similar pattern is followed throughout the development of the classifiers. In addition to standalone implementation of the classifiers, a GUI interface is developed for their comparison as well. The toolbox is tested on real world datasets to benchmarked against open source code snippets of the same classifiers. As the toolbox is developed with the sole purpose of convenience in mind, its target audience are researchers and instructors. The instructors can use this toolbox to teach a diverse range of classifiers. The researchers can use it to learn how classifiers work by employing them on their data. It can also be used to produce publishable results.
- Issued Date
- 2016
- Awarded Date
- 2016. 8
- Type
- Dissertation
- Alternative Author(s)
- Abbas, Waseem
- Department
- 대학원 해양시스템공학과
- Advisor
- 이종현
- Table Of Contents
- LIST OF FIGURES v
LIST OF TABLES v
ABSTRACT vii
ABBREVIATIONS AND NOTATIONS ix
Chapter 1 INTRODUCTION 1
1.1. Machine Learning . 1
1.2. Classification . 2
1.3. Related work . 4
1.4. Motivation . 6
1.5. Thesis Layout 8
Chapter 2 SHALLOW CLASSIFIERS . 10
2.1. Linear Classifiers 10
2.1.1. Fisher linear discriminant analysis 11
2.1.2. Least square classifier . 13
2.1.3. Principal component analysis (PCA) 13
2.1.4. Support vector machines (SVM) . 14
2.2. Probabilistic classifiers . 16
2.3. Neurons based classifiers 18
2.4. Chapter summary 20
Chapter 3 DEEP LEARNING CLASSIFIERS 22
3.1. Introduction 22
3.2. Multi-layer perceptron (MLP) . 23
3.2.1. Architecture . 23
3.2.2. General comments. 24
3.3. Convolutional Neural Network 24
3.3.1. Architecture . 24
3.3.2. Back Propagation 25
3.4. ELM based deep network 26
3.4. Deep belief networks (DBN) . 28
3.5. Chapter summary . 30
Chapter 4 IMPLEMENTATION . 31
4.1. Performance measure of classifiers 31
4.1.1. Precision 31
4.1.2. Recall 31
4.1.3. Accuracy . 31
4.1.4. ROC Curve 32
4.1.5. Confusion matrix. 32
4.2. Overview of implementation . 33
4.3. GUI development environment 34
4.4. GUIs of classifiers 35
4.4.1. Load Data 36
4.4.2. Parameters . 37
4.4.3. Train 45
4.4.4. Results . 46
4.5. Master GUI 47
4.5.1. Load Data 48
4.5.2. Parameters . 49
4.5.3. Train and Evaluate 49
4.5.4. Results . 49
4.5.5. Classifier Selection (For comparison) . 50
4.5.6. Comparison Results 51
4.6. Summary 51
Chapter 5 RESULTS OF THE TOOLBOX 52
5.1. Binary class classification 52
5.2. Multiclass classification . 54
Chapter 6 DISCUSSION AND CONCLUSION . 56
6.1. Discussion 56
6.1.1. GUI Support 56
6.1.2. Coherent Implementation 57
6.1.3. Modules Support . 57
6.1.4. Diversity of algorithms . 57
6.1.5. Visualization . 57
6.2. Conclusion and Future Work . 60
A 1 TYPOGRAPHIC CONVENTIONS AND DATA FORMAT 62
A 2 MATLAB FUNCTIONS USED . 65
Bibliography . 93
- Degree
- Master
- Publisher
- 제주대학교 대학원
- Citation
- Waseem Abbas. (2016). Matlab Based Machine Learning Software for Education
- Appears in Collections:
- Faculty of Earth and Marine Convergence > Ocean System
- 파일 목록
-
-
Download
Matlab Based Machine Learning Software for Education.pdf
기타 데이터 / 2.6 MB / Adobe PDF
-
Items in Repository are protected by copyright, with all rights reserved, unless otherwise indicated.