Machine Learning Resources

About this document

I compiled this list when transitioning out of academia into industry. My background is in (quantitative) economics and this list assumes that you know college-level math (calculus, linear algebra, probability/statistics) and some programming (e.g. Matlab).

Quick start

• Read ISL (free) for a gentle introduction to the theory behind important machine learning algorithms
• Read Hands on ML (Part I) to learn how to code up the most important algorithms in Python
• Take part in a Kaggle Challenge

Improving your coding skills

• Python
  • Data manipulation: Python for Data Analysis
  • Fundamentals: Learn Python the hard way (free)
  • Important libraries: numpy, pandas, seaborn, scikit-learn, keras, tensorflow
• R
  • Data manipulation and visualization: R for Data Science (free)
  • Fundamentals: Advanced R, Writing R Packages (free)
  • Important libraries: dplyr, ggplot2, tidyr, gbm, xgboost
• Version control
  • Happy git with R (free)
  • GitHub is your friend!

Deep Learning

• Goodfellow/Bengio/Courville (free) for theory
• Google’s (short) Tensorflow course on Udacity (free)
• Hands on ML (Part II) for implementation in using TensorFlow in Python

Visualization

• The classic for theory: The Visual Display of Quantitative Information
• Workhorse libraries are ggplot2 (R) and seaborn (Python)
• You can build quick, interactive dashboards in R using shiny and plotly. If you are serious about interactive visualization, then you may want to invest in D3

Archive

• For a statistical perspective on many ML techniques
  • Elements of Statistical Learning (free)
  • Computer Age Statistical Inference (free)
• For some of the more theoretical concepts in ML (e.g. VC-bound):
  • Learning from data
• Classics:
  • Pattern Recognition and Machine Learning
  • Machine Learning: A Probabilistic Perspective
• Online courses
  • Andrew Ng's Machine Learning course on Coursera
  • MIT 6.00.1 (EdX) for basic CS concepts with exercises in Python