2025 Activity Report

Published 2025-12-12


Introduction🔗

Our second year (January-December 2025) built on the first year's foundations. Results now include five published papers, a growing suite of open-source tools, and initial bridging from theory to brain-imaging data.

What we've done so far🔗

From two nodes to whole-brain connectomes🔗

Last year we mentioned studying simplified quadratic networks in the complex plane. This year we extended this analysis: for two-node networks, we characterized how the adjacency matrix and coupling weights determine whether the system converges to a stable attractor or exhibits chaotic dynamics. This research was consolidated in the paper "Asymptotic dynamics in systems of two coupled quadratic maps" and published in Chaos, Solitons & Fractals.

Furthermore, we scaled this up by introducing the equi-M set, a generalization of the classic Mandelbrot set for networks of any size. A key finding is that fractal and geometric invariants of these sets correlate robustly with the network's wiring. Applied to Human Connectome Project data, these invariants differentiate structural connectomes (anatomical wiring with only positive weights) from functional connectomes (activity-based correlations with both positive and negative weights), and distinguish between resting-state and emotion-processing task states. Notably, this method succeeds where classical graph-theoretic measures applied separately to the positive and negative subnetworks cannot. This work is part of the "Fractal geometry predicts dynamic differences in structural and functional connectomes" paper published in Chaos.

Wilson-Cowan with delays: Parkinson's disease and beyond🔗

We extended our stability and bifurcation analysis of Wilson-Cowan networks, first from 3-node networks to an arbitrary number of nodes. As a first step, we assumed identical delay distributions on all connections, which enabled a systematic analysis of how delays and connectivity jointly affect stability and bifurcations. We then generalized to the more realistic case where intra-node and inter-node connections have different delay distributions (two distinct delay kernels). We also systematically compared how the shape of the delay distribution (exponential, Gamma, Dirac, etc.) influences the linear spectrum and the onset of Hopf bifurcations. Even small perturbations to the delay profile can produce significant transitions between stationary behavior, periodic oscillations, and complex dynamics - an insight with direct relevance to the abnormal brain rhythms seen in Parkinson's disease.

This work was presented by E. Kaslik at NODYCON (2025), with two full papers in preparation for next year.

Generalized fractional operators and periodicity🔗

In related work, we explored generalized fractional operators based on Sonine kernels as an alternative to distributed-delay modeling. These operators have a Volterra-type representation, where the Sonine kernel acts mathematically as a delay distribution over the entire past interval. By choosing different kernel families (power-type, Prabhakar, distributed-order, etc.), one can model various long-term memory profiles.

In the paper "Generalized fractional operators do not preserve periodicity ", published in Fractional Calculus and Applied Analysis, we proved that these operators never preserve periodicity: for any non-constant periodic function, application of the operator yields an aperiodic function. The proof relies on a decomposition of the fractional derivative into a local periodic term and a nonlocal completely monotonic term responsible for the loss of periodicity. At the level of differential equations, this means fractional-order dynamical systems cannot admit exact periodic solutions - only asymptotically periodic ones. This is relevant for neural modeling because brain oscillations observed experimentally are generally only approximately periodic, influenced by slow adaptation and long-term memory processes. In future work, we plan to use generalized fractional operators with different Sonine kernels to construct fractional neural network models and compare the results with the distributed-delay approach.

Emotion, dopamine, and schizophrenia🔗

After an extensive revision based on reviewer feedback, our 4-dimensional delay-differential model of the prefrontal-limbic circuit in schizophrenia has been accepted for publication in Chaos, Solitons & Fractals (originally submitted in the first project stage). The paper shows how dopamine-modulated interactions in this emotion-regulation circuit can give rise to pathological dynamics - a result that connects theoretical analysis to clinical questions.

These findings were also presented by M. Neamțu at NODYCON (2025).

Software tools🔗

The software toolkit expanded this year:

These tools are open source and available on GitHub. For an up-to-date listing, see our Software page.

Empirical data analysis🔗

We also began exploring empirical data:

Both of these were presented at NODYCON 2025 by D. Jianu and S.-A. Folta, respectively.

Team activities🔗

Collaboration with our US partners at SUNY New Paltz (Anca Rădulescu) and University at Buffalo (Thomas J. Covey) was extensive. The Romanian team visited the US in June 2025; Anca Rădulescu visited Timișoara in July 2025. Weekly online meetings - in both extended teams and in focused subgroups for specific subprojects - maintained progress between visits.

Collectively the team gave presentations at three international conferences: NODYCON 2025 in Hoboken, ENUMATH 2025 in Heidelberg, and SYNASC 2025 in Timișoara.

By the numbers🔗

What we said we'd doWhat we did
3+ papers submitted5 published, 1 conference paper accepted
1 intermediate reportDone

The five journal papers appeared in Chaos, Solitons & Fractals (2), Fractional Calculus and Applied Analysis (2), and Chaos (1) - all Q1-ranked venues. Three of them are open access.

What's coming up🔗

With the analytical foundations solid and the software stack maturing, year three will focus on heterogeneous delays across network architectures, scaling the numerical simulations to larger networks, deepening the connection to empirical brain data, comparing fractional and distributed-delay approaches to neural modeling, and working toward integrating EEG and fMRI within a unified framework.