This blog will record the timeline, resources, projects along the way of my study on machine learning since 2015. And I will keep updating this blog.

Mathematics

• Basics: better to review before real study of ML
  • Calculus
  • Linear Algebra
    • Introduction to Linear Algebra: a nice textbook in linear algebra (First Pass in Dec, 2015).
    • MIT 18.06: Linear algebra (Completed in Dec, 2015).
  • Probability Theory
• Reference
  • Handbook of Mathematics: an amazing reference book of mathematics.
  • Problem-Solving Through Problems: Polish up your math skills.
• Statistics
  • All of Statistics: A textbook appealing to MLers (First Pass in Aug, 2017).
  • Statistical Inference: Classic textbook.
  • CMU 10-705: Intermediate Statistics (Completed in Aug, 2017).
  • Post Series: Doubt Clarification for Statistics (In Chinese)
    • 1. What is p-value?
    • 2. What’s the usage of probability inequalities
    • 3. Law of Large Numbers and Central Limit Theorem
• Optimization
  • Convex Optimization: Classic textbook. (First Pass in Nov, 2018)
  • Stanford EE364a: Convex Optimization. (Completed in Nov, 2018)
• Advanced Topics on Linear Algebra
  • The Matrix Cookbook: Ongoing
  • Matrix Analysis: Ongoing
• Advanced Topics on Statistics
  • A User’s Guide to Measure Theoretic Probability: Ongoing

Statistical Machine Learning

• Coursera: Machine Learning: Introduction to machine learning. (Certificate)
• StanfordOnline: Statistical Learning: Introduction to statistical learning and R. (Certificate)
• UAlberta CMPUT 466/551: Machine Learning. I’m one of TAs in this course.
  • Post Series: ML with R
    • 1. Linear Regression
    • 2. Regularized Linear Regression
    • 3. Logistic Regression
• The Elements of Statistical Learning: The first six chapters are highly recommended, while some later chapters are a little bit out-of-dated from my perspective (First Pass in Feb, 2018).
  • Manual: Figure out the mathematical details.
  • R package “ElemStatLearn”: Contain data sets, functions and examples from the book.
  • My code: Exercises and experiment reimplementations in R.
  • Post Series: Notes on Mathematics for ESL
    • 1. Chapter 2: Overview of Supervised Learning
    • 2. Chapter 3: Linear Regression Models and Least Squares
    • 3. Chapter 4: Linear Methods for Classification
    • 4. Chapter 5: Basis Expansions and Regularization
    • 5. Chpater 6: Kernel Smoothing Methods
    • 6. Chapter 10: Boosting and Additive Trees
• Machine Learning: A Probabilistic Perspective: In the view of Bayesian to interpret ML. (First Pass in Jan, 2019).

Probabilistic Graphical Models:

• UAlberta CMPUT 659: Probabilistic Graphical Models. Graduate course in University of Alberta based on Coursera course instructed by Koller (Report: A Topic Model of Genetic Mutations in Cancer, Completed in Apr, 2016).
• CMU 10-708: Probabilistic Graphical Models.
• Probabilistic Graphical Models: Principles and Techniques: Classic textbook for probabilistic graphical models. I’ve read a few chapters of it and gave up. This book is somewhat too mathematical and I will not recommend this book for beginners in PGM.
• Pomegranate: A Python package for PGM. Nice tutorials for beginners in PGM!

Deep Learning

• Neural Network and Deep Learning: Nice online tutorial for beginners in deep learning (Completed in Sep, 2016).
• Coursera: Neural Networks for Machine Learning: Excellent course for beginners in neural networks. (Certificate)
• Stanford CS224d: Deep Learning for Natural Language Processing (Completed in Apr, 2017).
• Stanford CS231n: Convolutional Neural Networks for Visual Recognition.
• Stanford CS236: Deep Generative Models. (Completed in Dec, 2018)
• Deep Learning Book: First and second part are great to read and need to be understood well. (First Pass in Oct, 2018)
• TensorFlow: Large amount of high-quality tutorials and well-coded recent advances in DL. Most applicable scenario: research.
• PyTorch: Use dynamic graph instead of static graph in TensorFlow. Relatively easy to use compared to TensorFlow. Most applicable scenario: research.
• Mxnet: High performance. Most applicable scenario: industry.
• Keras: High-level, easy to use. Most applicable scenario: competitions like Kaggle.

Note: I’m comfortable with coding in both TensorFlow and PyTorch and use Keras from time to time. I don’t know much about Mxnet and rarely use it, but I’m amazed by their group members.

Large Scale Machine Learning

• edX: Introduction to Big Data with Apache Spark: Introduction to Spark in Python. (Certificate)
• edX: Scalable Machine Learning: Explore deeper usage of Spark on large scale machine learning. (Certificate)
• Coursera: Functional Programming Principles in Scala: Introduction to Scala. (Certificate)
• Coursera: Functional Program Design in Scala: Further study on Scala. (Certificate)
• Coursera: Parallel Programming: Parallel computing in Scala. (Certificate)
• Coursera: Big Data Analysis with Scala and Spark: Intro to Spark with Scala. (Certificate)
• Coursera: Big Data Analysis with Scala and Spark: Intro to Spark with Scala. (Certificate)
• Coursera: Functional Programming in Scala Capstone: Capstone project for Scala. (Certificate)

Reinforcement Learning

• UAlberta CMPUT 609: Reinforcement Learning. Graduate course in UofA instructed by Sutton. (Report: Investigation on Deep Reinforcement Learning Methods for Classical Control Problems, Completed in Dec, 2016).
• Introduction to Reinforcement Learning: Classic textbook, recommended for beginners in RL (First Pass in Dec, 2016).
• RL Course from DeepMind: Instructed by David Silver. Highly recommended for beginners in RL (Completed in Oct, 2018)
• Bandit Algorithms Blog: A great blog discussing all kinds of bandit algorithms by Csaba (First Rough Pass in Feb, 2018).
• Post Series: Notes on Reinforcement Learning
  • 1. Finite Markov Decision Processes
  • 2. Dynamic Programming
  • 3. Monte Carlo Methods
  • 4. Temporal-Difference Learning

Related Topics

• Information Theory
  • Information Theory, Inference, and Learning Algorithm: Ongoing.
• Data Mining
  • Coursera: Mining of Massive Datasets: Introduction to a lot data mining techniques. (Certificate)
• Game Theory
  • Coursera: Game Theory: Introduction to game theory. (Certificate)

ML Competitions

• Kaggle: Home Depot Product Search Relevance: Top 10%. (Report)
• ATEC, NLP for Financial Integellience (hosted by Alibaba): 10th out of 443 teams (Certificate)