Computer Science Student at Université de Bordeaux interested in Artificial Intelligence, Machine Learning, Distributed Systems, Algorithmics.

This repository presents my academic and personal work, mainly focused on Artificial Intelligence and Theoretical Computer Science.

Each project includes a bibliography, they are all based on, at least, the following references:

• Géron, A. Machine Learning avec Scikit-Learn. Dunod.
• Charniak, E. Introduction au Deep Learning. Dunod.
• Géron, A. Deep Learning avec TensorFlow. Dunod.

Build a deep learning model to recognize cats from scratch, with emphasis on understanding the mathematical foundations.

• Multilayer Perceptrons (MLP)
  • Backpropagation algorithm
  • Activation functions and associated challenges
• Convolutional Neural Networks (CNN)
  • Feature map extraction and visualization
  • Pooling operations
• Custom Implementation
  • Built-from-scratch neural network model

Implement sequence-to-sequence models culminating in a functional French-to-English translator.

• Recurrent Neural Networks Fundamentals
  • Word embeddings
  • N-gram models (bigram, trigram)
• Advanced RNN Cells
  • Long Short-Term Memory (LSTM)
  • Gated Recurrent Units (GRU)
  • Text generation
• Sequence-to-Sequence Models
  • Encoder-decoder architecture
  • Attention mechanisms
  • Byte Pair Encoding (BPE) tokenization
• Final Project
  • French-to-English neural translator

Develop a deep reinforcement learning agent capable of learning to play Flappy Bird, inspired by DeepMind's 2013 Atari research.

• Reinforcement Learning Fundamentals
  • Core concepts and decision-making processes
  • Integration with deep neural networks
• Theoretical Foundations
  • Markov Decision Processes
  • Bellman equations
  • Policy gradient methods
• Deep RL Algorithms
  • Deep Q-Networks (DQN)
  • Double DQN architecture
  • Dueling networks
• Application
  • Flappy Bird learning agent

Generate realistic cat face images using generative models trained on a dataset of 30,000 64×64 cat face images.

• Autoencoders (AE)
  • Encoder-decoder architecture
  • Reconstruction loss and dimensionality reduction
• Variational Autoencoders (VAE)
  • Probabilistic latent space
  • Sampling and generation
• Generative Adversarial Networks (GAN)
  • Generator and discriminator dynamics
  • Comparative analysis with AE and VAE
• Implementation
  • Multi-generative model comparison
  • Generation from 30K cat face dataset (64×64)

Research project initiated during my internship at LaBRI, supervised by Mr. Raskin, a searcher at LaBRI.

The objective is to establish a lower bound for bidirectional enumeration problems under memory constraints, particularly for unidirectional data structures such as hash chains.

Academic work in Epistemology, nearing completion.
The project will be published as a public repository once finalized.

2nd Place Nuit de l’Info 2024
Challenge: "Mettez de l’intelligence dans votre moteur"
Organized by the Association Française pour l’Intelligence Artificielle (AFIA)

• link of the podium https://www.nuitdelinfo.com/nuitinfo/defis2024:archives#defi60
• link of the association : http://www.afia.asso.fr