Theme

This course given in English in the masters économie internationale et développement, digital economics and quantitative economic analysis. It gives a full coverage of modern machine learning.

Lectures notes/slides

Please note that the R scripts below have been extracted automatically from the knitr sources of the slides. They must be adapted to run properly: paths to data files must be modified and the opt_chunk related code must be removed. The code is developed under GNU/Linux and uses frequently the doMC package which is not available under MS Windows. It should be replaced by the doParallel package (and the code should be adapted).

Introduction

• slides on an introduction to machine learning: in English (R code). Last update: 02/10/2020.
• slides on an introduction to empirical risk minimisation: in English (last update: 22/01/2020)
• a short introduction to computational complexity: in English with R examples and English with python examples (last update: 01/03/2020)

Privacy

• slides on privacy: in English (last update: 06/07/2020)

Supervised learning models

• slides on decision trees: in English (R code). Last update: 22/01/2020.
• slides on optimal models and naive bayes: in English. Last update: 06/01/2020.
• slides on support vector machines and kernel methods: in English (R code). Last update: 06/01/2020.
• slides on ensemble methods (including random forests and boosting): in English (R code). Last update: 26/02/2020.

Tools for supervised learning

• slides on cross-validation and other resampling methods: in English (with R code), updated on 06/01/2020, and in French with more content (R code), updated on 10/01/2020.
• slides on imbalanced data: in French (R code), updated on 20/04/2020.

Unsupervised learning

• slides on clustering: English (R code). Last update: 18/10/2019.
• slides on outlier detection: English (R code). Last update: 16/06/2020.

Theory

• slides on empirical risk minimization: in English (last update: 01/24/2018)
• slides on regularization and capacity control: in English (last update: 01/24/2018)
• a more advanced and more thorough presentation of the same concepts are available in my slides on learning theory: in French and in English

Exercises and Labs

Introduction

• exercises on loss and risk (solutions)
• exercises on optimal constant predictions (solutions)

Decision trees

• labs on decision trees
  • in python
  • in R

Recommended reading/viewing

Full course

Tom Mitchell's and Nina Balcan's machine learning course:

• Introduction
• Decision Trees
• and basically all the video lectures

Selected topics

• Tom Mitchell's Naive Bayes Classifier primer