Source author record

Johan Nakuci

Johan Nakuci appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

nlin.CD Machine Learning Neurons and Cognition Social and Information Networks

Catalog footprint

What is connected

3works

4topics

4close collaborators

Actions

Connect this record

Open graph Browse works

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2022arXiv

Computing brain networks with complex dynamics

One important question in neuroscience is how global behavior in a brain network emerges from the interplay between network connectivity and the neural dynamics of individual nodes. To better understand this theoretical relationship, we have been exploring a simplified modeling approach in which we equip each node with discrete quadratic dynamics in the complex plane, and we study the emerging behavior of the resulting complex quadratic network (CQN). The long-term behavior of CQNs can be represented by asymptotic fractal sets with specific topological signatures going far beyond those described in traditional single map iterations. In this study, we illustrate how topological measures of these asymptotic sets can be used efficiently as comprehensive descriptors and classifiers of dynamics in tractography-derived connectomes for human subjects. We investigate to what extent the complex geometry of these sets is tied to network architecture (on one hand) and to the network behavior (on the other). This helps us understand the mechanics of the relationship between the subject's brain function, physiology and behavior and their underlying connectivity architecture.

preprint2022arXiv

Synchronization and clustering in complex quadratic networks

In continuation of prior work, we investigate ties between a network's connectivity and ensemble dynamics. This relationship is notoriously difficult to approach mathematically in natural, complex networks. In our work, we aim to understand it in a canonical framework, using complex quadratic node dynamics, coupled in networks which we call complex quadratic networks (CQNs). After previously defining extensions of the Mandelbrot and Julia sets for networks, we currently focus on the behavior of the node-wise projections of these sets, and on defining and analyzing the phenomena of node clustering and synchronization. We investigate the mechanisms that lead to nodes exhibiting identical or different Mandelbrot set. We propose that clustering is strongly determined by the network connectivity patterns, with the geometry of these clusters further controlled by the connection weights. We then illustrate the concept of synchronization in an existing set of whole brain, tractography-based networks obtained from 197 human subjects using diffusion tensor imaging. Synchronization and clustering are well-studied in the context of networks of oscillators, such as neural networks. Understanding the similarities to how these concepts apply to CQNs contributes to our understanding of universal principles in dynamic networks, and may help extend theoretical results to natural, complex systems.

preprint2020arXiv

Network Clustering Via Kernel-ARMA Modeling and the Grassmannian The Brain-Network Case

This paper introduces a clustering framework for networks with nodes annotated with time-series data. The framework addresses all types of network-clustering problems: State clustering, node clustering within states (a.k.a. topology identification or community detection), and even subnetwork-state-sequence identification/tracking. Via a bottom-up approach, features are first extracted from the raw nodal time-series data by kernel autoregressive-moving-average modeling to reveal non-linear dependencies and low-rank representations, and then mapped onto the Grassmann manifold (Grassmannian). All clustering tasks are performed by leveraging the underlying Riemannian geometry of the Grassmannian in a novel way. To validate the proposed framework, brain-network clustering is considered, where extensive numerical tests on synthetic and real functional magnetic resonance imaging (fMRI) data demonstrate that the advocated learning framework compares favorably versus several state-of-the-art clustering schemes.