# The computational role of structure in neural activity and connectivity

**Authors:** Srdjan Ostojic, Stefano Fusi  
**arXiv:** [2308.16772](https://arxiv.org/abs/2308.16772)  
**Date:** 2023-08-31  
**Summary page:** https://www.summarizepaper.com/en/paper/2308.16772v1/

## Key points

- Authors Srdjan Ostojic and Stefano Fusi address a significant challenge in neuroscience by identifying meaningful structures within chaotic neural activity.
- The proposed unified approach examines both representational geometry and modularity properties of recorded neural activity to unveil structural patterns within connectivity networks.
- Ostojic and Fusi establish a comprehensive framework for assessing geometry and modularity in both neural activity and connectivity, linking these properties to the computations carried out by neural networks.
- The authors highlight various types of structures identified in recent studies on model networks engaged in three distinct classes of computations, offering valuable insights into the interplay between structure, function, and computation in neural systems.

## Summary

In their paper "The Computational Role of Structure in Neural Activity and Connectivity," authors Srdjan Ostojic and Stefano Fusi address a significant challenge in neuroscience by identifying meaningful structures within chaotic neural activity. These structures hold key computational implications that shed light on the functional roles of specific brain regions. The authors propose a unified approach to characterizing these structures by examining both representational geometry and modularity properties of recorded neural activity. They demonstrate that this approach can unveil structural patterns within connectivity networks. To achieve this, Ostojic and Fusi establish a comprehensive framework for assessing geometry and modularity in both neural activity and connectivity. By linking these properties to the computations carried out by neural networks, they provide a deeper understanding of how structure influences function. Through their analysis, the authors highlight various types of structures identified in recent studies on model networks engaged in three distinct classes of computations. Overall, Ostojic and Fusi's work offers valuable insights into the intricate interplay between structure, function, and computation in neural systems. Their findings not only contribute to advancing our understanding of brain organization but also pave the way for future research aimed at deciphering the complexities of neural dynamics.