Lecturer: Matteo Neri
Fields: Computer Science, Robotics, Systems Engineering
Content
One of the central abilities of intelligent systems is the processing of information. In systems such as the human brain, large language models, or societies, intelligence arises from the interaction of many components, such as neurons in the brain or individuals in a society. Through their interactions, these components give rise to properties and functions that are often not predictable from the parts alone. It is precisely through these interactions that information is processed, and intelligence emerges (1).
A long tradition in information theory and graph theory has focused on the study of interactions, developing theoretical and computational tools to represent, describe, and infer from data how information is processed by systems composed of many interacting variables (2, 3). Over the past two decades, the increasing availability of data has enabled researchers to apply these tools to a wide range of complex systems, including the brain, deep neural networks, social systems, and financial markets (1).
In the first part of this presentation, I will provide a concrete illustration of the explanatory power of these approaches, focusing in particular on research strategies grounded in information- and graph-theoretic frameworks. I will refer to one of my early PhD papers (4, 8), in which I explore in detail the motivations and explanatory potential of this line of research. This work serves as a conceptual foundation for the subsequent sections of the talk.
I will then present the main outcomes of my research, which focus on the development of new computational tools and their application to the study of the human brain, social systems, and deep neural networks from an information- and graph-theoretic perspective (4, 5, 6, 7, 8). These approaches allow us to characterize how groups of units within a system interact to give rise to emergent information processing and integration. In particular, I will discuss results showing how different brain regions interact synergistically to integrate information during learning (5) and across different tasks (6). I will also present ongoing work investigating how similar information-dynamic principles may emerge across scales, from neural systems to social systems (manuscript in preparation).
Finally, I will conclude with a short hands-on data session. I will introduce a publicly available toolbox that enables students to apply the information- and graph-theoretic methods presented in the talk to their own data. This toolbox, which we developed and published in the Journal of Open Source Software (JOSS) during the early years of my PhD, is designed to be accessible even to users without extensive experience in Python (7). The goal of this final section is to allow participants to directly engage with the methods, explore their own scientific questions, and gain practical experience with these tools, providing a concrete and interactive conclusion to the presentation.
Literature
- 1. Varley, Thomas F. “Information theory for complex systems scientists.” arXiv preprint arXiv:2304.12482 (2023)
- 2. Shannon, Claude E. “A mathematical theory of communication.” The Bell system technical journal 27.3 (1948): 379-423.
- 3. Luppi, Andrea I., et al. “A synergistic core for human brain evolution and cognition.” Nature neuroscience 25.6 (2022): 771-782.
- 4. Neri, Matteo, et al. “A taxonomy of neuroscientific strategies based on interaction orders.” European Journal of Neuroscience 61.3 (2025): e16676.
- 5. Combrisson, Etienne, et al. “Higher-order and distributed synergistic functional interactions encode information gain in goal-directed learning.” Nature Communications 16.1 (2025): 7179.
- 6. Neri, Matteo et al. “Beyond Pairwise Interactions: Charting Higher-Order Models of Brain Function.” bioRxiv (2025): 2025-06.
- 7. Neri, Matteo, et al. “HOI: A Python toolbox for high-performance estimation of Higher-Order Interactions from multivariate data.” Journal of Open Source Software 9.103 (2024): 7360.
- 8. Neri Matteo, et al. “An integrated computational approach for diversity-sensitive personalized medicine.” Neuroscience (2025).
Lecturer
I hold a Bachelor’s degree in Mathematics and a Master’s degree in Complex Systems Physics. My main research interest is the investigation of the principles underlying consciousness and intelligence, which I am currently pursuing in a PhD program jointly conducted between Aix-Marseille University and Imperial College London. I have contributed to a total of 14 publications, including 6 as first author, and have presented my work at several international conferences and workshops. In parallel, I have teaching experience as a Teaching Assistant at Aix-Marseille University and as a mathematics tutor.
Affiliation: Institute de Neuroscience de la Timone
Homepage: https://mattehub.github.io/