Exercices

In this section, we propose quick exercises that can typically be done in an hour and a half.:

• Supervised learning with PyTorch: a toy example (duration: 1.5 h): Introductory exercise that will make you use a neural network modeled with PyTorch to classify points in supervised learning.

Practical work

In this section, we propose longer exercises that can be done without assistance, typically in 4 hours.

• Reinforcement learning on snake with a genetic neural network (duration: 4 h): An example of supervised learning applied to video games. You will use the legendary game Snake rewritten in Python for the occasion and will try to develop an automatic game strategy. In a first step, by hand and in a second step using a genetic algorithm to evolve a neural network. Graphical examples will allow to see the evolution of the game performances.