Coursebooks

Mathematical foundations of neural networks

MATH-631

Lecturer(s) :

Boumal Nicolas
Kressner Daniel
Nobile Fabio
Picasso Marco

Language:

English

Frequency

Only this year

Remark

This course addresses to students in mathematics and related fields, with a strong background and interest in mathematical aspects. Fall semester - Fridays from 18.09.2020

Summary

This course is in the form of a reading course / working group. We will focus on some mathematical aspects of the theory of neural networks, including universal approximation theorems, connections to ODEs and PDEs, optimiza-tion algorithms for NN training and their convergence.

Content

Goal:
Understanding the efficiency of deep neural networks in approximating functions in large dimensions and related problems

Content:
- Introduction to DNN
- Numerical experiments with available software
- Universal approximation theorems
- Connections to ODEs and PDEs
- Training of DNN

 

Note

A list of relevant research papers will be selected at the beginning of the course. Students will alternate in giving presentations on the selected papers.

Resources

Notes/Handbook

The list of scientific papers that will be studied during the course will be communicated on the first week.

In the programs

    • Semester
    • Exam form
       Oral presentation
    • Credits
      3
    • Subject examined
      Mathematical foundations of neural networks
    • Number of places
      24
    • Lecture
      16 Hour(s)
    • Practical work
      52 Hour(s)

Reference week

 
      Lecture
      Exercise, TP
      Project, other

legend

  • Autumn semester
  • Winter sessions
  • Spring semester
  • Summer sessions
  • Lecture in French
  • Lecture in English
  • Lecture in German