Sparse representations

the timeline of the whole course:

April 15th (morning+afternoon): basics on machine learning, practice with notebook using scikit learn (MG) April 16th (morning+afternoon): deep learning and automated differenciation, practice with notebook using pytorch (MG) April 17th morning: interpretable machine learning (ET) April 17th afternoon: sparse representations (LP)

If not done already, please install a (reasonably) recent version of python (easy option is anaconda, see details here: https://etulab.univ-amu.fr/gilson.m/compneuro_course). Importantly, part of the course will rely on pytorch, see instructions for installing a dedicated environment here: https://etulab.univ-amu.fr/gilson.m/compneuro_course/-/tree/main/autodiff (we can do together it the first morning for those who have trouble).

The first day (or morning depending on how we go), we will first review basics in supervised learning, to be on the same page (with a focus on recursive feature elimination): https://etulab.univ-amu.fr/gilson.m/compneuro_course/-/tree/main/sup_lrn If some of you are interested in machine learning for time series, we can have a session on this (we’ll decide together on Monday morning) Following, we will focus on autodifferenciation, first from scratch and then using pytorch, see https://etulab.univ-amu.fr/gilson.m/compneuro_course/-/tree/main/autodiff (in progress of being updated) And a few datasets are available there: https://etulab.univ-amu.fr/gilson.m/compneuro_course/-/tree/main/data ; in particular we will use the MNIST dataset as a benchmark for classification, etc.