Learning outcomes

At the end of the course, the student must demonstrate an understanding of the different topics covered (see content), i.e., be able to express in his own words the theory and methods seen in the course and explain in which context they are useful. He must also be able to implement the techniques seen during a complex data analysis problem.

Content

As an extension of the IDASM102 "Machine learning and data mining" course, this course explores more advanced methods in machine learning and deep learning.  The following topics will be discussed:

  1. probabilistic learning
  2. deep learning for images
  3. XAI and regularisation
  4. XGBoost and feature importance
  5. deep mearning for sequences
  6. density estimation and information theory
  7. generative deep learning models
  8. from text mining to large language models
  9. machine learning in the cloud and MLOps

These courses will be complemented by two "research talks" sessions based on live interventions by scientific experts and international conferences recorded and viewed during the lesson.  A session will also be devoted to the presentation of projects (see "evaluation mode").

Assessment method

The course is divided into two learning activities. The first consists of lectures and is evaluated by an oral examination on the theory of the course (70% of the overall course mark). The second is an ongoing assessment of students' ability to implement and document the techniques seen in class (30% of the overall course mark).  To do this, he will have to carry out a project whose approximate timing is:

  1. End of September: introduction session to the usual frameworks in deep learning
  2. Early October: project kick-off session by the assistants
  3. End of October: Q&A session / project progress checkpoint
  4. Mid-November: presentation of the results obtained

The project will be evaluated on the basis of regular submissions on an online platform and the final presentation.  It will be carried out in teams of two students.

Sources, references and any support material

References are given during the course.

 

Language of instruction

Anglais