Teaching

Lectures, labs and supervision

Lectures & labs

Télécom Paris

Graph Mining

  • 2025 - 2026
  • 130 students
  • 2 lectures · 3 labs
  • 7.5 h

Learn the main properties of real graphs, such as scale-free and small-world structure, and key algorithms to sample, rank, classify, represent and cluster nodes.

  • Course 1: sparse matrices and the general properties of large graphs.
  • Course 7: Graph Neural Networks, from message passing to GCN and GraphSAGE.
Course 1: Graph structure
Introduction PDF Sparse matrices PDF General properties of large graphs PDF
Course 7: Graph Neural Networks
Slides PDF Lecture notes PDF
Mirakl

Graph Machine Learning

  • 2026
  • Data scientists and a general audience
  • 2 lectures
  • 4 h

Graph ML explained at two levels, a technical one for data scientists and a high-level one for everyone, co-written with Robin Brochier.

  • Patterns of large graphs: sparsity, power laws, small world.
  • PageRank, to find the most important nodes.
  • Graph Neural Networks: from message passing to their limitations and extensions.
  • Applications of GNNs, and opportunities for Mirakl.
Slides
For data scientists: a gentle introduction to GNNs PDF For everyone: Graph ML PDF

Supervision

Télécom Paris

Data Science & AI project: Better Embeddings for Better GNNs

  • Apr - Jun 2026
  • 6 students
  • 7-week project

The team compared text features (TF-IDF, BoW, sBERT, CLIP-text, Qwen3) and image embeddings, alone and combined, to measure how node features affect a GraphSAGE model on Amazon e-commerce co-purchase graphs.

Deliverable
Project report PDF
Mirakl

Internship: GNN product recommender

  • Apr - Sep 2026
  • Neil Braun
  • 6-month internship

Supervised Neil Braun while he built a product recommender system based on Graph Neural Networks.