Summary
A fully funded 3-year PhD position is available under the Hi! PARIS project SIGMA at LTCI (Télécom Paris) and SAMOVAR (Télécom SudParis), Institut Polytechnique de Paris. The research focuses on graph signal processing and geometric deep learning, specifically aiming to jointly compress graph structures and their associated signals while preserving essential information for reconstruction and downstream graph neural network (GNN) tasks.
PhD Scholarship: Signal-Aware Graph Summarization with GNN Guarantees (SIGMA) France
Designation
PhD Researcher / PhD Candidate
Key Information at a Glance
| Parameter | Details |
| Project Title | Signal-aware Graph Summarization with GNN Guarantees (SIGMA) |
| Position Type | PhD Position (Full-Time) |
| Duration & Start | 3 Years, starting December 2026 |
| Funding | Funded by Hi! PARIS |
| Supervisors | Dr. Jhony H. Giraldo & Dr. Aref Einizade |
| Host Institutions | LTCI (Télécom Paris) & SAMOVAR (Télécom SudParis), Institut Polytechnique de Paris |
| Location | Paris-Saclay cluster, France |
| Application Deadline | Rolling basis (Open until filled) |
Research Area
- Primary Domains: Graph Signal Processing, Graph Machine Learning, Geometric Deep Learning.
- Keywords: Graph summarization, graph signals, higher-order structures, graph neural networks (GNNs), scalable learning.
Location
Télécom Paris & Télécom SudParis, Paris-Saclay cluster, France (located on the outskirts of Paris, ~45 minutes by train from Paris city center).
Eligibility / Qualification
Candidates must meet the following criteria:
- Academic Background: Currently holding or finishing a Master 2 (M2) or equivalent degree in Engineering, Data Science, Computer Science, Applied Mathematics, Signal Processing, Statistics, or related fields.
- Core Competencies: Strong background in signal processing and machine learning, with a genuine interest in graph signal processing and geometric deep learning.
- Programming Skills: High proficiency in Python, including experience with PyTorch.
- Theoretical Foundation: A strong interest in understanding the underlying mathematics behind graph signal processing and geometric deep learning (a firm requirement).
- Soft Skills: Clear communication skills.
Job Description
Large-scale graphs in social networks, recommender systems, biological networks, and knowledge graphs contain millions or billions of nodes and edges, posing severe computational and storage challenges.
This PhD project focuses on developing unified compression methods that jointly summarize graph topology and graph signals into a compact representation.
- Primary Focus: Investigate methods to compress graphs and their associated signals together while preserving key details for signal reconstruction and machine learning tasks.
- Methodology: Utilize meaningful higher-order structures as units of summarization to capture complex relationships beyond simple node-and-edge pairs.
- Scope: Explore when these higher-order structures are most effective and develop efficient selection algorithms for scalable application on massive real-world graphs.
How to Apply
To apply, send your complete application package via email directly to the project supervisors:
- Jhony H. Giraldo:
jhony.giraldo@telecom-paris.fr - Aref Einizade:
aref.einizade@telecom-sudparis.eu
Required Application Materials:
- Full CV.
- Motivation Letter explaining your interest in the position (maximum 1 page).
- Transcript of records (grades).
- At least one letter of recommendation.
Last Date to Apply
- No Fixed Deadline (Rolling Basis): Applications are reviewed as they are received, and the position will close as soon as a suitable candidate is selected. Early application is strongly encouraged.
