CV

Research interests

Graph theory, combinatorics, graph machine learning, expressivity of graph neural networks, fairness in machine learning, user modeling.

Education

• Ph.D. candidate, Télécom Paris, Paris, France, 2024 – 2027 (defense planned for January 2027)
  • Predicting Edges, Constraining Edges: From Fairness in Link Prediction to Edge-Girth Realizability
  • Advisors: Charlotte Laclau, Tiphaine Viard, Bruno Sguerra
  • Research area: graph theory, combinatorics, machine learning, algorithmic fairness
• Master’s degree MVA (Mathematics, Vision, Learning), ENS Paris-Saclay, Paris, France, 2022 – 2023
  • Relevant courses: convex optimization, kernel methods for machine learning, deep learning, time series analysis, image denoising, introduction to digital imaging, deep learning in practice
• Engineering degree, ENSAE Paris (Data Science, Statistics and Learning), Palaiseau, France, 2019 – 2023
  • Relevant courses: Bayesian statistics, machine learning theory, reinforcement learning, Monte Carlo methods, optimization

Internships

• Deezer Research, Paris, France, May – November 2023
  • Music taste modeling: user embeddings based on listening regularities
  • Encoding user activity into time series, pattern detection through dictionary learning
  • Supervisors: Bruno Sguerra, Manuel Moussallam
• Louis Bachelier Institute, Paris, France, September 2021 – March 2022
  • Natural language processing: sentence embeddings, sentiment analysis, topic modeling
• CETIC, Charleroi, Belgium, May – August 2021
  • Hybrid CNN and LSTM neural networks for the detection of stenoses in coronary arteries from X-ray images
  • Supervisor: Xavier Lessage

Publications

Structural Bias Beyond Homophily: A Study of Fairness in Link Prediction

Lilian Marey, Mathilde Perez, Tiphaine Viard, Charlotte Laclau. Structural Bias Beyond Homophily: A Study of Fairness in Link Prediction. Transactions on Machine Learning Research (TMLR), 2026.

Edge-Girth as a Structural Edge Feature for Graph Neural Networks

Lilian Marey, Charlotte Laclau. Edge-Girth as a Structural Edge Feature for Graph Neural Networks. NeurIPS 2026 Workshop: Bridging Optimal Transport, Learning and Structured Data: Toward Geometric Distributional Learning (non-archival), 2026.

On the Realizability of Edge-Girth Sequences

Lilian Marey, Paul Hilaire, Charlotte Laclau. On the Realizability of Edge-Girth Sequences. Preprint, arXiv:2607.25629, 2026.

k-hop Fairness: Addressing Disparities in Graph Link Prediction Beyond First-Order Neighborhoods Permalink

Lilian Marey, Tiphaine Viard, Charlotte Laclau. k-hop Fairness: Addressing Disparities in Graph Link Prediction Beyond First-Order Neighborhoods. Preprint, arXiv:2603.03867, 2026.

Modeling Musical Genre Trajectories through Pathlet Learning

Lilian Marey, Charlotte Laclau, Bruno Sguerra, Tiphaine Viard, Manuel Moussallam. Modeling Musical Genre Trajectories through Pathlet Learning. Adjunct Proceedings of the 33rd ACM Conference on User Modeling, Adaptation and Personalization (UMAP), 2025, pp. 202–210.

Do Recommender Systems Promote Local Music? A Reproducibility Study Using Music Streaming Data

Kristina Matrosova, Lilian Marey, Guillaume Salha-Galvan, Thomas Louail, Olivier Bodini, Manuel Moussallam. Do Recommender Systems Promote Local Music? A Reproducibility Study Using Music Streaming Data. Proceedings of the 18th ACM Conference on Recommender Systems (RecSys), 2024, pp. 148–157.

Modeling Activity-Driven Music Listening with PACE

Lilian Marey, Bruno Sguerra, Manuel Moussallam. Modeling Activity-Driven Music Listening with PACE. Proceedings of the 2024 ACM SIGIR Conference on Human Information Interaction and Retrieval (CHIIR), 2024, pp. 346–351.

A New Coronary Artery Stenosis Detection Method with a Hybrid LSTM-CNN Model

Xavier Lessage, Michal Nedoszytko, Saïd Mahmoudi, Lilian Marey, Olivier Debauche, Sidi Ahmed Mahmoudi. A New Coronary Artery Stenosis Detection Method with a Hybrid LSTM-CNN Model. Artificial Intelligence and Smart Environment (ICAISE 2022), Lecture Notes in Networks and Systems, Springer, pp. 70–77.

Presentations

Edge-Girth as a Structural Edge Feature for Graph Neural Networks

2026 (upcoming)

Poster at NeurIPS 2026 Workshop: Bridging Optimal Transport, Learning and Structured Data: Toward Geometric Distributional Learning, Paris, France

Beyond Dyadic Fairness: Distance-Aware Bias in Graph Link Prediction

2026

Poster at Fifth European Conference on Algorithmic Fairness (ECAF 2026), Ghent, Belgium

Beyond Dyadic Fairness: Distance-Aware Bias in Graph Link Prediction

2026

Poster at Conférence sur l'Apprentissage automatique (CAp 2026), Montpellier, France

Modeling Activity-Driven Music Listening with PACE

2024

Poster at ACM SIGIR Conference on Human Information Interaction and Retrieval (CHIIR 2024), Sheffield, UK

Teaching

Artishow projects (first-year engineering students)

Machine Learning for Text Mining (Master’s Degree)

Statistical learning and data mining (Master’s Degree)

Skills

• Languages: French, English, Spanish
• Computational tools: exhaustive graph generation and verification of constructions (networkx, nauty)
• Programming: Python (numpy, PyTorch, scikit-learn, pandas)