Intro 2 AI - Fall 2026 class
Note: The course is given in-person (“presentiel”) at CerCo, Pavillon Baudot, Hopital Purpan (Salle de Conference 127, 1st floor).
Schedule (click on each lecture for details)
| Topic | Lecture # | Teacher | Sub-topic | Date | Time |
|---|---|---|---|---|---|
| Neural Networks | Lecture 1 | Jan Bellingrath | Neural Networks: History and Foundations | Sept 7 | 1.30-3.30pm |
| Computer Vision | Lecture 2 | Bastien Lelan | Image classification | Sept 7 | 4-6pm |
| Computer Vision | Lecture 3 | Bastien Lelan | Unsupervised/zero/few-shot learning | Sept 14 | 1.30-3.30pm |
| Computer Vision | Lecture 4 | Jan Bellingrath | Object detection, segmentation | Sept 14 | 4-6pm |
| NLP | Lecture 5 | Raja Kumar | Natural Language Processing basics | Sept 21 | 1.30-3.30pm |
| NLP | Lecture 6 | Raja Kumar | Recurrent Neural Networks for NLP | Sept 21 | 4-6pm |
| NLP | Lecture 7 | Raja Kumar | Transformers for NLP & LLMs | Sept 28 | 1.30-3.30pm |
| Computer Vision | Lecture 8 | Luca Gonzalez-Sommer | Multimodal Learning | Sept 28 | 4-6pm |
| Audio | Lecture 9 | Jean-Felix Maestrati | Sound processing, speech recognition | Oct 5 | 1.30-3.30pm |
| DL/Neuro | Lecture 10 | Rufin VanRullen | Homologies between brain & CNNs | Oct 5 | 4-6pm |
| DL/Neuro | Lecture 11a | Alexandre Queant | Spiking Neural Networks | Oct 19 | 2-3pm |
| DL/Neuro | Lecture 11b | Luca Gonzalez-Sommer | Brain decoding with Machine Learning | Oct 19 | 3.15-4.15pm |
Audience
This class, provided by ANITI, is intended for a M2-level audience with no advanced background in Computer Science or Maths (e.g. Neuro, Bio, Medical, etc.). The classes will be in English, and there will be a mixture of lectures and practical work (TP), with at-home assignments. For the practical parts, prior experience with Python programming will be required.
Do you need a refresher on the basics of Python programming before starting our class? To find out, look at [this online course], from lecture “Datatype and Variables” to lecture “Python object and classes” (included). There is nothing to install, all the code examples can be run interactively in the browser. If you already know all the concepts in this course, then your Python level is sufficient. If you’re a beginner, it might take you 10+ hours to go through the course, but this will be necessary to follow our class…
Other useful resources include [this course], or [this one].
Similarly, if you’d like to brush up math concepts that are relevant to modern AI, you could check [this blog] or [this one], or [this two-hour video] or [this one].
Grading
Grading will be based on individual projects. Each student should choose ONE of the course notebooks (from lectures 1-9) and complement it with a significant personal work. Just create one or more cells at the end of the notebook, clearly marked with your personal addition. It could be a change in the model’s architecture, an adaptation of the model to a new task or dataset, etc. If you lack inspiration, feel free to start from the “exercises” or personal work suggested at the end of each notebook (but projects with original and creative thinking will get extra points). The link to the colab notebook (after making sure that it is “shared for anyone with the link”) should be submitted by email to rufin.vanrullen@cnrs.fr on or before Nov 15. Do NOT submit via moodle. This is a firm deadline, the grade will decrease by 1 point for each late day (e.g. if you send your notebook on Nov 25, the maximum grade you can get is 10/20). The main criteria for grading will be:
- The addition should be clearly marked in the notebook, e.g. with a title “MY PERSONAL ADDITION IN THE 3 CELLS BELOW”
- The addition should be explained (e.g. with comments in the code), so one can at least know what you intended to achieve
- The addition should be non-trivial
- The addition should reflect significant work and effort
- The addition should make it clear that the student has understood the lecture and notebook
- The addition should work, i.e. it should be possible to evaluate the cells without returning an error.