Objectifs :
The purpose of this course is to provide the student with the mathematical framework and technical toolkit needed in order to understand advanced topics in machine learning and be able to adapt and refine reference methodologies to meet their scientific needs.
Given its brevity, the course does not attempt to provide an exhaustive overview of the wide variety of ML sub-disciplines, but to give the student the tools to autonomously delve deeper in these subjects without losing track of their scientific aim.

Programme :
1 - Complexity and other core concepts

2 - Linearity is not what it seems

3 - Building neurons

4 - SVMs and kernels

5 - Generating with Bayes

6 - The curse of dimensionality

7 - Clustering

8 - Trees, forest, and wise crowds

Pré-requis :
The course assumes a good expertise in Python programming (up and included the use of classes and objects), as well as a good understanding of multivariate calculus, linear algebra, and probability. Topics such as – but not limited to - partial derivatives, matrix diagonalization, expectation values will be used with little to no introduction. A self-evaluation test will be available for the student to decide whether or not take this course.

Equipe pédagogique :
- Edoardo SARTI - Ercan SECKIN

Méthode pédagogique :
Cours magistraux et TD

Compétences acquises à l'issue de la formation :
- Clearly explain both the logic and the mathematical framework behind some of the most renown ML methodologies

- Detail the mathematical derivation of some ML algorithms

- Use off-the-shelf implementations of ML algorithms in diverse biomedical contexts

- Correct, modify, adapt, refine existing ML algorithms to meet specific scientific needs

- Analyze, preprocess, and transform data for ensuring an effective and correct use of a target ML method

La formation participe à l'objectif suivant :conforter la culture scientifique des doctorants dans leur champ disciplinaire ou en interdisciplinaire