Deep learning et applicationsEA Big Data et IA (STI)

This course aims to acquire the theoretical and practical foundations of deep learning, to understand the main architectures of neural networks and their fields of application. At the end of the course, students will be able to design, train and evaluate deep learning models, exploit pre-entry models and develop applications incorporating large-scale language models (LLMs).

Deep learning foundations: neural networks, activation functions, gradient backpropagation, optimization and regulation.

Neuronal architectures: convolutive networks (CNN), recurring networks (RNN, LSTM), applications to vision and sequence processing.

Transform architectures: attention and self-attention mechanisms, positional encoding, encoder-decoder architectures.

LLM and IA generative: foundation models, pre-training, fine tuning, text generation and model adaptation.

Advanced applications: rapid engineering, increased generation by recovery (RAG), use of tools and agent systems (Agentic AI).

Implementation: development of applications with Python and PyTorch/Hugging Face; performance evaluation and model boundary analysis.

Continuous monitoring