Deep-Learning


Generalities

Mini Project

Objectives:
For the Project you will have to implement a model based on one of the following paper. I recommend for image based model to use a small dataset and not the dataset proposed in the paper (to limit computationnal resources):
  • (CycleGAN) Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks
  • (VAE) DRAW: A Recurrent Neural Network For Image Generation
  • (Attention based model) End-To-End Memory Networks
  • (diffusion model) Denoising Diffusion Probabilistic Models
Also you can choose lower hidden size to ensure it runs on your computer
Submission
You should send a pdf report using science article template ( ICLR template). You should:
  • Introduce your topic (with a research question) -- up to 1 page
  • Discuss related works -- up to 1 page
  • Describe the method -- up to 1 page
  • Present your experimental setup (evaluation metrics, dataset, parameters of the model, ...) -- up to 1.5 pages
  • Report and discuss the result of your experiments -- up to 2 pages
  • Conclude and discuss possible improvments/limitations -- up to .5 page
Deadline
April 30 2026

Session 1: Introduction

  • Introduction to Deep-Learning (Slides)

Session 2: Gradient Descent and Backpropagation

Session 3: Backpropagation and Optimisation

Session 4: Pytorch and Introduction to CNN

Session 5-6: VAE and exercises

Submission of the labs is expected April 23rd.