topology-dependent-polychronous-neuronal-groups

name	topology-dependent-polychronous-neuronal-groups
description	Topology-Dependent Emergence of Polychronous Neuronal Groups using Recurrence-Plot characterization. Small-world topology as structural optimum for polychronization with label-free PNG identification via sparse-dot-product Recurrence Plot framework.
version	1
author	extracted from arXiv 2606.25874v1 (Carneiro, Jiofack, Ferreira)
date_created	2026-06-25T00:00:00.000Z
source	arXiv 2606.25874v1
categories	["neuroscience","spiking-neural-networks","computational-neuroscience","neural-dynamics","brain-network"]
tags	["polychronous-neuronal-groups","PNG","recurrence-plot","STDP","axonal-delays","Izhikevich-neurons","watts-strogatz","small-world-topology","network-structure"]
activation_keywords	["polychronous neuronal groups","PNG emergence","recurrence plot neural networks","topology neural computation","STDP polychronization","axonal delays PNG","small-world neural network","Izhikevich PNG","clustering coefficient neural","recurrence quantification analysis neuroscience"]
related_skills	["spiking-neural-network-analysis","stdp-spiking-transformer-attention","spiking-oscillation-mapping","brain-inspired-snn-pattern-analysis"]

Topology-Dependent Emergence of Polychronous Neuronal Groups: A Recurrence-Plot Characterization

Overview

Core Discovery: Small-world topology is the structural optimum for polychronization in neural networks. The clustering coefficient C is the primary structural driver of Polychronous Neuronal Group (PNG) yield, with transition from ring-lattice to random graph reducing representational capacity by >90%.

Methodological Innovation: Sparse-dot-product Recurrence Plot (RP) framework for label-free PNG identification — identifies PNGs as unit-slope diagonal structures in phase-space recurrence matrix, independent of anatomical neuron labeling.

Key Findings

1. PNG Structural Determinants

• N=1000 Izhikevich neurons simulated over 10 hours biological time
• 1545 unique PNGs identified via offline event-driven detection
• Clustering coefficient C drives PNG yield: C~0.35 (ring-lattice) → 850 PNGs; C0.20 (random) → <50 PNGs
• >90% representational capacity reduction from structured to random topology

2. Small-World Topology as Optimum

• Ring-lattice (high clustering) supports rich PNG repertoire
• Random graphs (low clustering) drastically reduce PNG diversity
• Small-world networks optimize trade-off between local structure and global connectivity

3. Recurrence Plot PNG Decoder

• Sparse-dot-product RP identifies PNGs without neuron labeling
• PNGs appear as unit-slope diagonal structures in phase-space recurrence matrix
• Recurrence Quantification Analysis: DET~0.65 quantifies trajectory reproducibility
• Label-free, principled approach to PNG identification

Technical Implementation

Izhikevich Network Simulation

# Izhikevich neuron model parameters
# N = 1000 neurons
# STDP plasticity with heterogeneous axonal delays
# Watts-Strogatz topology sweep: p ∈ [0, 1]
# Clustering coefficient C as structural metric

# PNG detection algorithm (event-driven, offline)
# 1. Track spike timings across all neurons
# 2. Identify time-locked firing cascades
# 3. Verify reproducibility via STDP stabilization