Graph Neural Networks for Neuroheuristic Brain Analysis

UNIL principal investigator

Prof. Alessandro Villa, Faculty of Business and Economics

UNIPD principal investigator

Prof. Alessandro Sperduti, DM


Joint seminar / conference involving early-stage researchers


With this project proposal, we aim to start a collaboration between the NeuroHeuristic Research Group (HEC-ISI) at the University of Lausanne, and the Machine Learning Group (MLG) and Computational Cognitive Neuroscience Lab (CCNL) at the University of Padua.

In light of the expertise of these three groups, our goal is to investigate how Graph Neural Networks can be applied to brain data to analyze and understand the pattern of connectivity in the biological neural system that accounts for human behavior.

Our intention is to promote and coordinate this collaboration by organizing two series of seminars for participants of the project from the MLG/CCNL groups and the NeuroHeuristic Research Group. The goal of the first series of seminars will be to lay the groundwork for the joint project. In the second series of seminars, the researchers will present and discuss the early results.

We expect that the project will result in publishing a paper at a top-ranking international conference (and possibly a follow-up journal article) and implementing public outreach activities designed to demonstrate the potential for an academic transdisciplinary project to spark scientific innovation with applications throughout a wide range of disciplines.


The main objective of the project is to start a collaboration between the NeuroHeuristic Research Group (HEC-ISI) at the University of Lausanne, and the Machine Learning Group (MLG) and Computational Cognitive Neuroscience Lab (CCNL) at the University of Padua.

The HEC-ISI studies complex neural mechanisms relying on the neuroheuristic approach. Such an integrative approach develops along an experimental axis based on complex behavioral paradigms that involve learning and memory phenomena occurring not only in animals but also in humans. The MLG has an extensive expertise in machine learning for graph-structured data, including the design and application of graph kernels and graph neural networks. The CCNL is a multidisciplinary research group investigating the neural mechanisms underlying human cognition through computational models and neuroinformatics tools.

The overarching goal of this project is to enhance our understanding of the brain as a complex self-organizing system, how its topological properties drive the interplay between sensory processing, sensorimotor integration, and cognition, and how these properties are affected by brain diseases. In our expectation, the interdisciplinary approach and collaboration between HEC-IS, MLG, and CCNL would lead to results that will have a broad impact  across many areas of neuroscience research and to the fast-growing application of Artificial Intelligence in predictive medicine.

We foresee two collaborative events among project participants:

  • The first one will take place in the last quarter of 2022. This first event concerns a series of seminaries involving the participants of the projects from the NeuroHeuristic Research Group, and two early-stage researchers from the MLG/CCNL groups. The aim of these seminars is to lay the foundations for the joint project.
  • A second event will take place in the first quarter of 2023. This time the event will be hosted by the department of Mathematics of the University of Padova. In these second series of seminars, the researchers will present and discuss the early results of the joint project, and study the next steps to be taken for achieving the goals of the project.

During the whole project, online meetings will take place every month.

Finally, we plan to complete the project and submit a paper to an international conference before September 2023 and to a top-tier peer-reviewed journal at the earliest completion date. The paper will summarize the main research results obtained during the project.

Potential for follow-up activities

Machine learning models have been applied, in the last few years, to several problems in medicine, chemistry, biology and many others. While such techniques cannot replace the capabilities of human experts, for sure it can provide valuable tools to ease many activities, and it can be used to provide useful insights for future research. In this project, the aim is to extract new knowledge analyzing brain activity patterns with ML models for graphs, that can extract complex patterns from the available data that can be difficult to identify for a human.

If the goal of this project would be achieved, it would pave the way for novel approaches to the neuro heuristic brain analysis that may allow for a better understanding of the dynamics of human brain activity while performing higher cognitive tasks, such as decision-making tasks.

As an output of this project, we expect to publish at least one paper in a top-ranking international conference and implement public outreach activities to illustrate the potential of an academic transdisciplinary project for a scientific innovation with applications in a broad range of disciplines. Moreover, we plan to extend the planned conference proceeding into a full paper to be submitted to a high-impact open access international peer-reviewed journal. Since the project duration is relatively short, we did not include any budget to cover the expense for open access publication. However, the project partners will cover such publication costs with external funding and support of the university academic services.

Suivez nous: