Grass-roots optimization of coupled oscillator networks

Synchronization is critical for system function in applications ranging from cardiac pacemakers to power grids. Existing optimization techniques rely largely on global information, and while they induce certain local properties, those alone do not yield optimal systems. Therefore, while useful for designing man-made systems, existing theory provides limited insight into self-optimization of naturally-occurring systems that rely on local information and offer limited potential for decentralized optimization. Here we present a method for grass-roots optimization of synchronization, which is a multiscale mechanism involving local optimizations of smaller subsystems that are coordinated to collectively optimize an entire system, and the dynamics of such systems are particularly robust to islanding or targeted attacks. In addition to shedding light on self-optimization in natural systems, grass-roots optimization can also support the parallelizable and scalable engineering of man-made systems.

Evolutionary Reinforcement Learning Dynamics with Irreducible Environmental Uncertainty

In this work we derive and present evolutionary reinforcement learning dynamics in which the agents are irreducibly uncertain about the current state of the environment. We evaluate the dynamics across different classes of partially observable agent-environment systems and find that irreducible environmental uncertainty can lead to better learning outcomes faster, stabilize the learning process and overcome social dilemmas. However, as expected, we do also find that partial observability may cause worse learning outcomes, for example, in the form of a catastrophic limit cycle. Compared to fully observant agents, learning with irreducible environmental uncertainty often requires more exploration and less weight on future rewards to obtain the best learning outcomes. Furthermore, we find a range of dynamical effects induced by partial observability, e.g., a critical slowing down of the learning processes between reward regimes and the separation of the learning dynamics into fast and slow directions. The presented dynamics are a practical tool for researchers in biology, social science and machine learning to systematically investigate the evolutionary effects of environmental uncertainty.

Bursting in a next generation neural mass model with synaptic dynamics: a slow-fast approach

We report a detailed analysis on the emergence of bursting in a recently developed neural mass model that takes short-term synaptic plasticity into account. The one being used here is particularly important, as it represents an exact meanfield limit of synaptically coupled quadratic integrate & fire neurons, a canonical model for type I excitability. In absence of synaptic dynamics, a periodic external current with a slow frequency {epsilon} can lead to burst-like dynamics. The firing patterns can be understood using techniques of singular perturbation theory, specifically slow-fast dissection. In the model with synaptic dynamics the separation of timescales leads to a variety of slow-fast phenomena and their role for bursting is rendered inordinately more intricate. Canards are one of the main slow-fast elements on the route to bursting. They describe trajectories evolving nearby otherwise repelling locally invariant sets of the system and are found in the transition region from subthreshold dynamics to bursting. For values of the timescale separation nearby the singular limit {epsilon} = 0, we report peculiar jump-on canards, which block a continuous transition to bursting. In the biologically more plausible regime of {epsilon} this transition becomes continuous and bursts emerge via consecutive spike-adding transitions. The onset of bursting is of complex nature and involves mixed-type like torus canards, which form the very first spikes of the burst and revolve nearby fast-subsystem repelling limit cycles. We provide numerical evidence for the same mechanisms to be responsible for the emergence of bursting in the quadratic integrate & fire network with plastic synapses. The main conclusions apply for the network, owing to the exactness of the meanfield limit.

Selection of activity patterns in multiplex neuronal networks

We present a study on the selection of a variety of activity patterns among neurons that are connected in multiplex framework, with neurons on two layers with different functional couplings. With Hindmarsh-Rose model for the dynamics of single neurons, we analyze the possible patterns of dynamics in each layer separately, and report emergent patterns of activity like anti-phase oscillations in multi-clusters with phase regularities and enhanced amplitude and frequency with mixed mode oscillations when the connections are inhibitory. When they are multiplexed with neurons of one layer coupled with excitatory synaptic coupling and neurons of the other layer coupled with inhibitory synaptic coupling, we observe transfer or selection of interesting patterns of collective behaviour between the layers, inducing anti-phase oscillations and multi-cluster oscillations. While the revival of oscillations occurs in the layer with excitatory coupling, the transition from anti-phase to in-phase and vice versa is observed in the other layer with inhibitory synaptic coupling. We also discuss how the selection of these patterns can be controlled by tuning the intra-layer or inter-layer coupling strengths or increasing the range of non-local coupling. With one layer having electrical coupling while the other synaptic coupling of excitatory(inhibitory)type, we find in-phase(anti-phase) synchronized patterns of activity among neurons in both layers.

Length Regulation Drives Self-Organization in Filament-Motor Mixtures

Cytoskeletal networks form complex intracellular structures. Here we investigate a minimal model for filament-motor mixtures in which motors act as depolymerases and thereby regulate filament length. Combining agent-based simulations and hydrodynamic equations, we show that resource-limited length regulation drives the formation of filament clusters despite the absence of mechanical interactions between filaments. Even though the orientation of individual remains fixed, collective filament orientation emerges in the clusters, aligned orthogonal to their interfaces.

Dynamical-systems theory of cellular reprogramming

In cellular reprogramming, almost all epigenetic memories of differentiated cells are erased by the overexpression of few genes, regaining pluripotency, potentiality for differentiation. Considering the interplay between oscillatory gene expression and slower epigenetic modifications, such reprogramming is perceived as an unintuitive, global attraction to the unstable manifold of a saddle, which represents pluripotency. The universality of this scheme is confirmed by the repressilator model, and by gene regulatory networks randomly generated and those extracted from embryonic stem cells.

Compositional Active Inference I: Bayesian Lenses. Statistical Games

We introduce the concepts of Bayesian lens, characterizing the bidirectional structure of exact Bayesian inference, and statistical game, formalizing the optimization objectives of approximate inference problems. We prove that Bayesian

Homeostatic Criticality in Neuronal Networks

In self-organized criticality (SOC) models, as well as in standard phase transitions, criticality is only present for vanishing driving external fields $h rightarrow 0$. Considering that this is rarely the case for natural systems, such a restriction poses a challenge to the explanatory power of these models. Besides that, in models of dissipative systems like earthquakes, forest fires and neuronal networks, there is no true critical behavior, as expressed in clean power laws obeying finite-size scaling, but a scenario called dirty criticality or self-organized quasi-criticality (SOqC). Here, we propose simple homeostatic mechanisms which promote self-organization of coupling strengths, gains, and firing thresholds in neuronal networks. We show that near criticality can be reached and sustained even in the presence of external inputs because the firing thresholds adapt to and cancel the inputs, a phenomenon similar to perfect adaptation in sensory systems. Similar mechanisms can be proposed for the couplings and local thresholds in spin systems and cellular automata, which could lead to applications in earthquake, forest fire, stellar flare, voting and epidemic modeling.

Secondary frequency control stabilizing voltage dynamics

The ongoing energy transition challenges the stability of the electrical power system. Stable operation of the electrical power grid requires both the voltage (amplitude) and the frequency to stay within operational bounds. While much research has focused on frequency dynamics and stability, the voltage dynamics has been neglected. Here, we study frequency and voltage stability in the case of the simplest network (two nodes) and an extended all-to-all network via linear stability and bulk analysis. In particular, our linear stability analysis of the network shows that the frequency secondary control guarantees the stability of a particular electric network. Even more interesting, while we only consider secondary frequency control, we observe a stabilizing effect on the voltage dynamics, especially in our numerical bulk analysis.

Low dimensional behaviour of generalized Kuramoto model

We study the global bifurcations of frequency weighted Kuramoto model in low-dimension for network of fully connected oscillators. To study the effect of non-zero-centered frequency distribution, we consider two symmetric Lorentzians as an example. We derive the stability diagram of the system and show that the infinite-dimensional problem reduces to a flow in four dimensions. Using the system symmetries, it can be further reduced to two dimensions. Using this analytic framework, we obtain bifurcation boundaries of the system, which is compatible with our numeric simulations. We show that the system has three types of transitions to synchronized state for different parameters of the frequency distribution: (1) a two-step transition, representative of standing waves, (2) a continuous transition, as in the classical Kuramoto model, and (3) a first-order transition with hysteresis. Numerical simulations are also conducted to confirm analytic results.