The Neural Graph Inverse Problem

The project sets out to build the Neural Graph Inverse Problem framework, a common “toolbox” that can reverse-engineer hidden connections within many kinds of network-shaped data — whether those networks come from biological processes, social media interactions, or supply chain optimization problems. By treating each task as a puzzle of working backward from observed data to the underlying network that produced it, we aim to deliver a shared family of methods, accessible benchmarking procedures, and open-source code. This unified approach will help researchers avoid working in isolated silos, speed up progress across different fields, and make results easier to reuse. The Mila — Quebec AI Institute (Canada) team will primarily contribute through its strength in data-driven machine-learning applications, while the host team in RWTH Aachen (Germany) brings deep theoretical know-how in graph algorithms; together we will co-develop new techniques, publish joint research, and train students who can bridge theory and practice across both institutions.

Faculty Supervisor:

Guy Wolf

Student:

Partner:

Rheinisch-Westfälische Technische Hochschule Aachen

Discipline:

Mathematics

Sector:

Education

University:

Université de Montréal

Program: