README.md

Machine Learning Practice

Some practices using statistical machine learning technique based on some dataset.

To see more detail or example about deep learning, you can checkout my Deep Learning repository.

Because Github don't support LaTeX for now, you can use the Google Chrome extension TeX All the Things (github) to read the notes.

Environment

• Using Python 3

(most of the relative path links are according to the repository root)

Dependencies

• numpy: For low-level math operations
• pandas: For data manipulation
• sklearn - Scikit Learn: For evaluation metrics, some data preprocessing

For comparison purpose

• sklearn: For machine learning models
• cvxopt: For convex optimization problem (for SVM)
• For gradient boosting
  • XGBoost
  • LightGBM
  • CatBoost

For visualization

• Mlxtend
• matplotlib
  • matplotlib.pyplot
  • mpl_toolkits.mplot3d

For evaluation

• surprise: A Python scikit building and analyzing recommender systems

NLP related

• gensim: Topic Modelling
• hmmlearn: Hidden Markov Models in Python, with scikit-learn like API
• jieba: Chinese text segementation library
• pyHanLP: Chinese NLP library (Python API)
• nltk: Natural Language Toolkit

Projects

Subject	Technique / Task	Dataset	Solution	Notes
Letter Recognition	kNN / Classification	Letter Recognition Datasets (File)	kNN From Scratch, kNN Scikit Learn	Notes
Page Blocks Classification	Decision Tree / Classification	Page Blocks Classification Data Set (File)	Decision Tree (CART) From Scratch, Decision Tree Scikit Learn	Notes
CSM	Linear Regression / Regression	CSM Dataset (2014 and 2015) (File)	Linear Regression From Scratch, Linear Regression Scikit Learn, Linear Regression PyTorch NN	Notes
Nursery	Naive Bayes / Classification	Nursery Data Set (File)	Gaussian Naive Bayes From Scratch, Gaussian Naive Bayes Scikit Learn	Notes
Post-Operative Patient	SVM (cvxopt) / Binary Classification	Post-Operative Patient Data Set (File, Simplified)	SVM From Scratch (using cvxopt and simplified dataset), SVM Scikit Learn	Notes
Student Performance	AdaBoost / Classification	Student Performance Data Set (File)	AdaBoost From Scratch, AdaBoost Scikit Learn	Notes
Sales Transactions	k-Means / Clustering	Sales Transactions Dataset Weekly (File)	k-Means From Scratch, k-Means Scikit Learn	Notes
Frequent Itemset Mining	FP-Growth / Frequent Itemsets Mining	Retail Market Basket Data Set (File)	FP-Growth From Scratch	Notes
Automobile	PCA / Dimensionality Reduction	Automobile Data Set (File)	PCA From Scratch, PCA Scikit Learn	Notes
Anonymous Microsoft Web Data	SVD / Recommendation System	Anonymous Microsoft Web Data Data Set (File, Ratings Matrix (by R))	SVD From Scratch, R Notebook - IBCF Recommender System	Notes
Handwriting Digit	SVM (SMO) / Binary & Multi-class Classification	MNIST (File)	Binary SVM From Scratch, Multi-class (OVR) SVM From Scratch	Notes
Chinese Text Segmentation	HMM (EM) / Text Segmentation & POS Tagging	File	HMM From Scratch, HMM hmmlearn, Compare with Jieba and HanLP	-
Document Similarity and LSI	VSM, SVD / LSI	Corpus of the People's Daily (File)	VSM From Scratch, VSM Gensim, SVD/LSI Gensim	Notes
Click and Conversion Prediction	Logistic Regression / Recommendation System	Ali-CCP (File too large about 20GB)		Notes
LightGBM & XGBoost & CatBoost Practice	Boosting Tree / Classification	Social Network Ads (File)	LightGBM, XGBoost	Notes
Kaggle Elo	LightGBM / Feature Engineering	Elo Merchant Category Recommendation	LightGBM Project
DCIC 2019	LXGBoost / Feature Engineering	Failure Prediction of Concrete Piston for Concrete Pump Vehicles	XGBoost Project
Epinions CLiMF	Collaborative Filtering / Recommendation System	Epinions	CLiMF From Scratch, CLiMF TensorFlow	Notes, PaperResearch
Iris EM	EM Algorithm / Clustering	Iris Data Set	EM From Scratch	Notes
Iris Logistic	Logistic Regression / Classification	Iris Data Set	Logistic Regression From Scratch, Logistic Regression Scikit Learn, SVM (used for compare)	Notes

Machine Learning Categories

Consider the learning task

• Surpervised Learning
  • Classification - Discrete
  • Regression - Continuous
• Unsupervised Learning
  • Clustering - Discrete
  • Dimensionality Reduction - Continuous
  • Association Rule Learning
• Semi-supervised Learning
  • Semi-Clustering
  • Semi-Classification
• Reinforcement Learning

Consider the learning model

• Discriminative Model
  • Discriminative Function
  • Probabilistic Discriminative Model
• Generative Model

Cosider the desired output of a ML system

• Classification
  • Logistic Regression (optimization algo.)
    • Multinomial/Softmax Regression (SMR)
  • k-Nearest Neighbors (kNN)
  • Support Vector Machine (SVM) - Derivation (optimization algo.)
  • Naive Bayes
  • Decision Tree (ID3, C4.5, CART)
• Regression
  • Linear Regression - Derivation (optimization algo.)
  • Tree (CART)
• Clustering
  • k-Means
  • Hierarchical Clustering
  • DBSCAN
• Association Rule Learning
  • Apriori
  • Eclat
  • FP-growth - Frequent itemsets mining
• Dimensionality Reduction
  • Principal Compnent Analysis (PCA)
  • Single Value Decomposition (SVD) - LSA, LSI, Recommendation System
  • Canonical Correlation Analysis (CCA)
  • Isomap (nonlinear)
  • Locally Linear Embedding (LLE) (nonlinear)
  • Laplancian Eigenmaps (nonlinear)

Ensemble Method (Meta-algorithm)

• Bagging
  • Random Forests
• Boosting
  • AdaBoost <- With some basic boosting notes
  • Gradient Boosting
    • Gradient Boosting Decision Tree (GBDT) (aka. Multiple Additive Regression Tree (MART))
  • XGBoost
  • LightGBM

NLP Related

• Hidden Markov Model (HMM) - Sequencial Labeling Problem
• Conditional Random Field (CRF) - Classification Problem (e.g. Sentiment Analysis)

Backbone

• Maximum Entropy Model (MEM)
• Bayesian Network (aka. Probabilistic Directed Acyclic Graphical Model)

Others

• Probabilistic Latent Semantic Analysis (PLSA)
• Latent Dirichlet Allocation (LDA)
• Vector Space Model (VSM)
• Radial Basic Function (RBF) Network
• Isolation Forest
  • paper
• One-Class SVM

Heuristic Algorithm (Optimization Method)

• SMO --> SVM
• EM --> HMM, etc.
• GIS == improved ==> IIS --> MEM

Machine Learning Concepts

General Case

• Data Preprocessing
  • Normalization
  • Training and Test Sets - Splitting Data
  • Missing Value
  • Dimensionality Reduction
  • Feature Scaling
• Model Expansion
  • Binary to Multi-class
• Fitting and Model Complexity
  • Overfitting
  • Underfitting
  • Generalization
  • Regularization
• Reducing Loss
  • Learning Rate
  • Gradient Descent
• Other Learning Method
  • Cost-sensitive Learning
  • Lazy Learning
  • Incremental Learning (Online Learning)
  • Multi-label Classification

Categorized

• Classification
  • Data Preprocessing
    • Label Encoding
  • Real-world Problem
    • Cost-sensitive Learning
    • Classification Imbalance
  • Evaluation Metrics
    • Classification Metrics
  • Binary to Multi-class Expension
• Regression
  • Evaluation Metrics
    • Regression Metrics
• Clustering
  • Evaluation Metrics
    • Clustering Metrics

Specific Field

• Data Mining - Knowledge Discovering
• Feature Engineering
  • Training optimization
    • Memory usage
    • Evaluation time complexity
• Recommendation System
  • Collaborative Filtering (CF)
• Information Retrieval - Topic Modelling
  • Latent Semantic Analysis (LSA/LSI/SVD)
  • Latent Dirichlet Allocation (LDA)
  • Random Projections (RP)
  • Hierarchical Dirichlet Process (HDP)
  • word2vec

Machine Learning Mathematics

Topic

• Kernel Usages
• Convex Optimization
• Distance/Similarity Measurement - basis of clustering and recommendation system

Categories

• Linear Algebra
  • Orthogonality
  • Eigenvalues
  • Hessian Matrix
  • Quadratic Form
  • Markov Chain - HMM
• Calculus
  • Multivariable Deratives
    • Quadratic Approximations
    • Lagrange Multipliers and Constrained Optimization - SVM SMO
    • Lagrange Duality
• Probability and Statistics
  • Statistical Estimation
    • Maximum Likelihood Estimation (MLE)

Basics

• Algebra
• Trigonometry

Application

(from A to Z)

• Decision Tree
  • Entropy
• HMM
  • Markov Chain
• Naive Bayes
  • Bayes' Theorem
• PCA
  • Orthogonal Transformations
  • Eigenvalues
• SVD
  • Eigenvalues
• SVM
  • Convex Optimization
  • Constrained Optimization
  • Lagrange Multipliers
  • Kernel

Books Recommendation

Machine Learning

• Machine Learning in Action
  • Source Code
• 統計學習方法 (李航)
• 機器學習 (周志華) (alias 西瓜書)
  • datawhalechina/pumpkin-book - 南瓜書
    • github
• Python Machine Learning
  • Source Code
• Introduction to Machine Learning 3rd
  • Solution Manual
  • Previous version: 1st, 2nd
• Automated Machine Learning: Methods, Systems, Challenges (AutoML)
  • pdf

Mathematics

• Linear Algebra with Applications (Steven Leon)
• Convex Optimization (Stephen Boyd & Lieven Vandenberghe)
• Numerical Linear Algebra (L. Trefethen & D. Bau III)

Resources

Tutorial

Videos

• Google - Machine Learning Recipes with Josh Gordon
  • Josh Gordon's Repository
• Youtube - Machine Learning Fun and Easy
• Siraj Raval - The Math of Intelligence
  • Siraj Raval's Repository
• bilibili - 機器學習 - 白板推導系列
• bilibili - 機器學習升級版

Documentations

• ApacheCN (ML, DL, NLP)
  • Github - AiLearning
  • Official Website - ApacheCN 中文開源組織
• Machine learning 101 (infographics)
  • Machine learning overview (infographic)
  • Machine learning methods (infographic)
  • Machine learning evolution (infographic)
• Machine Learning Cheatsheet

Interactive Learning

• Google Machine Learning Crash Course
• Learn with Google AI
• Kaggle Learn Machine Learning
• Microsoft Professional Program - Artificial Intelligence track
• Intel AI Developer Program - AI Courses
  • Introduction to AI
  • Machine Learning
  • Time Series Analysis

MOOC

• Stanford Andrew Ng - CS229
  • Coursera

Github

• Machine Learning from Scratch (eriklindernoren/ML-From-Scratch)
• Avik-Jain/100-Days-Of-ML-Code - 100 Days of ML Coding
  • Chinese version
• ddbourgin/numpy-ml - Machine learning, in numpy
• Machine learning Resources
• microsoft/recommenders - Best Practices on Recommendation Systems
• dformoso/machine-learning-mindmap
  • pdf

Textbook Implementation

• Machine Learning in Action
  • Jack-Cherish/Machine-Learning
• Learning From Data (林軒田)
  • Doraemonzzz/Learning-from-data
• 統計學習方法 (李航)
  • WenDesi/lihang_book_algorithm
    • blog
  • fengdu78/lihang-code
  • wzyonggege/statistical-learning-method
  • Dod-o/Statistical-Learning-Method_Code
    • blog
  • SmirkCao/Lihang
• Stanford Andrew Ng CS229
  • fengdu78/Coursera-ML-AndrewNg-Notes
  • fengdu78/deeplearning_ai_books
• NTU Hung-Yi Lee
  • LeeML-Notes

Datasets

• UCI Machine Learning Repository
• Awesome Public Datasets
• Kaggle Datasets
• The MNIST Database of handwritten digits
• 資料集平台 Data Market
• AI Challenger Datasets
• Peking University Open Research Data
• Open Images Dataset
  • github
• Alibaba Cloud Tianchi Data Lab
• biendata

Competition

Global

• Kaggle Competition
• KDDCup
  • KDD Cup 2019: AutoML for Temporal Relational Data
    • 4Paradigm
    • CodaLab
• CodaLab

Taiwan

• Open Data

China

• Tianchi Competition
• biendata
• SODA
• Data Fountain
• Data Castle

Machine Learning Platform

• Kaggle
• OpenML

Machine Learning Tool

• AutoML
  • Auto-sklearn
    • github
• Optuna - A hyperparameter optimization framework
  • github
• Hyperopt - Distributed Asynchronous Hyper-parameter Optimization
  • github
  • hyperopt-sklearn

(Online) Development Environment

• Google Colaboratory
• Kaggle Kernels
• Intel AI DevCloud

jupyter notebook

• Extension plugin - pip install jupyter_contrib_nbextensions
  • VIM binding
  • Codefolding
  • ExecuteTime
  • Notify
• Jupyter Theme - pip install --upgrade jupyterthemes