Differences

This shows you the differences between two versions of the page.

--- curriculum [2025/06/18 10:30] – [Transverse Courses and Projects (spanning Period 1 and 2)] respai-vic
+++ curriculum [2025/06/23 11:42] (current) – respai-vic
@@ Line 74: / Line 74: @@
-**CSC_53432_EP - Large Language Models (24h, 2 ECTS), Guokan Shang (EP) ** (contact: guokan.shang@polytechnique.edu)
+**CSC_53432_EP - Large Language Models (24h, 2 ECTS), Guokan Shang (MBZUAI) ** (contact: guokan.shang@mbzuai.ac.ae)
-> This course offers an extensive and in-depth exploration of natural language processing (NLP) and large language models (LLMs), blending foundational principles with advanced techniques. The curriculum spans the domain of NLP, starting from basic concepts like indexing, Bag-of-Words, and TF-IDF, and extending to more sophisticated methods such as Graph of Words, Word2Vec, fastText, and GloVe. Then the focus shifts to deep learning models in NLP, including 1D-CNN, RNN, Hierarchical Attention Networks (HAN), and Neural Machine Translation (NMT).  Additionally, It delves into essential evaluation metrics and benchmark. Following that, the course presents the cutting-edge transformer architecture and pretraining techniques such as masked and causal language modelling. Key models including BERT, GPT, and BART are studied in detail, highlighting their transformative impact on the field and all their applications. The course also covers the emergence of LLMs introduced by ChatGPT, emphasizing prompting techniques, model adaptation using methods like LoRA, and model quantization techniques including techniques like QLoRA that merges both model adaptation and quantization. This comprehensive course integrates hands-on lab sessions using PyTorch and Keras, ensuring students gain practical experience alongside theoretical knowledge, preparing them for advanced roles in data science and NLP.
+> This course offers a deep dive into Large Language Models (LLMs), blending essential theory with hands-on labs to develop both practical skills and conceptual understanding—preparing you for roles in LLM development and deployment.
+> The curriculum begins with a brief overview of key historical NLP techniques. It then transitions to the transformer architecture, focusing on its attention mechanism and tokenization—the core of modern LLMs. Pre-training objectives such as masked/denoising language modeling and causal language modeling will also be covered, forming the basis for models like BERT, GPT, and T5. The course then examines LLM post-training techniques used to refine pre-trained models, including instruction tuning (SFT), reinforcement learning from human feedback (e.g., PPO/DPO), and reinforcement learning from verifiable rewards (e.g., GRPO). Finally, the course will address LLM application and future directions—including RAG, agents, multimodality, and alternative model architectures.
 **CSC_53439_EP - Deep Reinforcement Learning (24h, 2 ECTS), Jesse Read** (contact: jesse.read@polytechnique.edu)
@@ Line 102: / Line 103: @@
 **CSC_54434_EP - 3D Computer Vision (24h, 2 ECTS), Xi Wang (EP)** (contact: Xi.Wang@polytechnique.edu)
-> (ABSTRACT TO BE ADDED)
+> This course presents modern 3D computer vision in a clear, step-by-step progression: we begin with classical multi-view reconstruction and structure‑from‑motion pipelines, then advance to neural implicit representations for novel‑view synthesis (e.g., NeRF), proceed to explicit geometry rendering via 3D Gaussian Splatting (3DGS), and finally explore generative 3D models: e.g., 3D generation guided by 2D generative score distillation (DreamFusion-like), data‑driven 3D/4D content generation for dynamic scenes and motion, and the latest video generation techniques.
 **CSC_54443_EP - Soft robots: Design, Modeling, Simulation and Control (24h, 2 ECTS),  Christian Duriez (Inria Lille)** (contact: christian.duriez@inria.fr)