Selected by Sorbonne University within the framework of the Congé pour Projet Pédagogique (CPP), Philippe Esling (STMS-Ircam) will develop an innovative pedagogical program in AI that is in line with the institution's strategy, the national strategy in AI and the international context to train more students in modern artificial intelligence techniques.

Presentation 
The aim of the project is to develop a unified pedagogical toolkit that takes advantage of the most recent technologies best suited to AI learning, based on
on a model of cooperative and spiral pedagogy. Thus, the project intends to develop around the Jupyter Notebook technology by setting up remote learning servers, thus avoiding the problems of set-up and installation. 

This technology will also make it possible to offer simultaneous distance learning. In addition, the development of the tools will be done by using the Reveal-JS and RISE3 systems allowing to mix masterful presentation and code evaluation. This will make it easier to bring together theoretical and practical aspects by allowing students to follow the presentation (slides), while being able to modify its behaviour interactively. This aspect of bringing theoretical and practical aspects closer together is also at the heart of the project, which aims to develop a set of pedagogical notions in cycles of 1h30. Finally, the learning time aspects of the models will be mitigated by the development of toy datasets (allowing for rapid learning) as well as semi-trained models, which can be used interactively, thanks to the use of interactive systems similar to those proposed by Distill and based on systems such as D3.js.

 

Teaching Objectives
- New interactive learning model combining lecture, tutorials and practical work in a single class, using 1.5 hour cycles of pedagogical concepts.
- Implementation of a unified learning environment
- Use of the latest technologies in interactive learning
- Model based on iPython notebook / jupyterlab
- Implementation of a continuous learning server (jupyterhub / nbviewer) - Hybrid presentation using RISE / Jupyter
- Creation of widgets and interactive examples (distill.pub)
- Interactive graph rendering (D3.js - MPLD3 // Bokeh)
- Use of simplification extensions (Jupyter-contrib extensions)