Winter-Summer schools

Occasionaly I lecture for winter/summer schools. Here are the materials for the last one.

• Lecture slides
• Lecture 1 video
• Lecture 2 video

Data Analysis (Info)

This course is divided into 5 chapters and aims at providing the basic element for any statistical analysis. Consequently we will talk about descriptive statistics, data visualization, inferential statistics, clustering, principal component analysis and linear models. As this lecture is given to informatic guys, most of the theoretical development will not be covered. Here is the lecture material

Statistics (BioSTIC)

This lecture is divided into two parts. The first one is about statistical learning and will cover unsupervised classification using k-means, principal component analysis and logistic regression. The second part is about survival analysis and will talk about specificities of survival data, non parametric estimation of survival curves, comparison of survival curves and Cox's proportional hazard model. The lecture materials for this course are here (part 1, part 2). Labs sheet are listed below:

• TP1 k-means in practive with an application on Fisher's iris and italian wine datasets.
• TP2 Using PCA on a socio-economical (thanks André Mas!) and football players
• TP3 On the use of logistic regression on the survial on the Titanic and survival for cancer.
• TP4 Survival analysis with R.

Statistics 2 (MathAppli)

This course is divided in 3 independent parts:

• Non parametric statistics: lectures, Exercise sheet
• Bayesian statistics: lecture (annotated slides), Exercise sheet
• Time series: lecture, Exercise sheet

Here are (some) solutions to the exercises done during our lectures.

Bayesian Statistics and Hierarchical Models (MathAppli)

This course is an 'advanced' lecture on Bayesian statistics with an emphasis on computational statistics for Bayesian studies. During the lecture we will:

• briefly review the core concepts in Bayesian inference;
• introduce MCMC samplers (Metropolis--Hastings and Gibbs);
• talk about (Bayesian) hierarchical models;
• briefly talk about mixture models.
• intrudoce approximate Bayesian computation.
• apply methodologies to various applications.

You can get the lecture slides (annotated version) and empty document for questions. The pieces of code written during the lecture are given below:

• The hard-shell challenge
• The dumb student challenge
• The repulsive point example
• Your first Gibbs algorithms
• The football model (version html)
• The Bioasay
• The Coagulation time analyssis (Bayesian one-way anova)

GRADING GRADING GRADING

To grade follow the instructions

Advanced statistical learning (MathAppli)

While Bertrand Michel taught various statistical learning techniques, I will focus on two main themes: reinforcement learning and model based clustering. The lecture materials are available below:

1. Reinforcement learning: lecture notes and annotated version
• Lab 0: Introduction to the Gymnasium framework (Notebook)
• Lab 1: The Self driving taxi (q-learning) (Notebook)
• Lab 2: The cart pole challenge (REINFORCE) (Notebook)
2. Model based clustering: lecture notes and annotated version
• Lab 3: Gaussian mixtures (Notebook)
• Lab 4: Some applications of Gaussian mixtures (Notebook)

Statistical Learning (FastTrack)

This course is a (kind of) mini course on statistical learning. During the lecture (syllabus might change depending on students levels), we will:

• motivate the use a statistical learning techniques and introduce the basic elements (loss, risk, cross validation...)
• briefly review the core concept in statistics
• introduce the linear and logistic regression
• talk about regularized linear regression (ridge, lasso and elastic net)
• review basic theory about gradient-based optimization
• investigate (deep) neural networks (mostly feed forward and convolutional)

You can get the lecture slides (annotated version). The labs are listed below:

• Lab 0: Introduction to R (Rmd)
• Lab 1: Linear models (Rmd)
• Lab 2: Logistic regression (Rmd)
• Lab 3: Regularized regression (Rmd)
• Lab 4: Neural Network and Deep learning (Rmd)

Some code written during lectures.

• On the use of MLE with application
• Modelling air quality using linear models (webpage)
• Invest money using logistic regression (webpage)

Survival analysis

This course is a (rather short) course on survival analysis. During the lecture (syllabus might change depending on students levels), we will:

• talk about the specificities of survival analysis (skewness, censoring, ...)
• introduce the main theoretical quantities of interest (survivorship, hazard rate, ...)
• talk about non parametric estimation (Kaplan-Meier, Nelson-Alen, ...)
• test for difference in survivorship (log-rank, wilcoxon, ...)
• talk about semi-parametric regression (Cox's proportional hazards)

You can get the lecture slides here (annotated version) and hand written notes.

Extreme value theory (Toulouse)

This course is an introduction to the univariate extreme value theory with a view towards financial and environmental applications. You can download the slides (annotated ones). I list below some data sets used during the lecture and the lab sessions:

• Daily precipitation dataset at Toulouse Blagnac.
• Toulouse daily minimum temperature dataset at Toulouse Blagnac.
• Toulouse daily maximum temperature dataset at Toulouse Blagnac.
• Here is the Apple data set (Rdata format).

Statistics (STAD)

This course is a (kind of) overall introduction on some statistical topics. The course is divided into two parts: survival analysis and classical statistical models. During the lecture (syllabus might change depending on students levels), we will:

• talk about the specificities of survival analysis (skewness, censoring, ...)
• introduce the main theoretical quantities of interest (survivorship, hazard rate, ...)
• talk about non parametric estimation (Kaplan-Meier, Nelson-Alen, ...)
• test for difference in survivorship (log-rank, wilcoxon, ...)
• Principal component analysis
• Unsupervised classification with K-means
• Linear models and logistic regression
• More stuffs???

You can get the lecture slides here: survival analysis and and general statistics (annotated slides).

Here are some datasets that we used during the lectures:

• Socio-economic
• French Football team from Fifa22 (RData file)

Here are some code:

• mixing PCA and k-means

GRADING GRADING GRADING
Here are the instructions for grading the survival analysis part and that for the statistical learning part.

Data Analysis (Ecole Centrale Casablanca)

This course is divided into 2 independent chapters. The first one is about Geostatistics where you will learn basic concepts such as gaussian random fields, variogram fitting, kriging and (conditional) simulations. The second chapter is about machine learning where CART, random forest and boosting will be covered. All lecture materials (including Labs) are given below:

• Lectures materials (annotated slides)
• Labs:
  • Geostatistics (html, Rmd)
  • Random Forests (notebook)
  • Boosting (notebook)