A Structured Neuron-Space for Analysing Functional Organization in Spiking Networks

Authors: Mehran Jaberinezhad

Presentation type: Poster at SNUFA 2026 online workshop (4-5 Nov 2026)

Abstract

Understanding how individual neurons contribute to computation remains challenging in recurrent spiking neural networks, where neuronal properties, activity patterns, and connectivity jointly determine network function. Existing analyses often characterize these aspects separately, making it difficult to obtain a unified representation of individual neurons and to examine how their functional relationships emerge within a trained circuit.

Here, we investigate a structured representation in which each neuron is described by a multidimensional state incorporating properties of its intrinsic dynamics, activity, and interactions with the surrounding network. Rather than treating these measurements as independent descriptors, we use the resulting representation as a space in which relationships between neurons can be studied geometrically.

We develop a computational framework for constructing and analyzing this neuron-level representation in recurrent spiking networks and investigate how its structure relates to functional organization. In particular, we examine whether geometric relationships between neurons capture similarities in their activity and computational roles, and how these relationships change as networks learn a task. We further explore whether the resulting representation can reveal organization that is not apparent from individual firing-rate or connectivity statistics alone.

This approach provides a potential framework for connecting microscopic neuronal properties with the organization and function of trained spiking circuits. More broadly, it may offer a way to study learning-induced changes in neural representations at the level of individual neurons rather than only at the population level.

Are you the author and do you want to add some related links to this page? If so, check out our guide to adding related links and videos.