Pulse-coupled systems such as spiking neural networks exhibit nontrivial invariant sets in the form of attracting yet unstable saddle periodic orbits where units are synchronized into groups. Heteroclinic connections between such orbits may in principle support switching processes in these networks and enable novel kinds of neural computations. For small networks of coupled oscillators, we here investigate under which conditions and how system symmetry enforces or forbids certain switching transitions that may be induced by perturbations. For networks of five oscillators, we derive explicit transition rules that for two cluster symmetries deviate from those known from oscillators coupled continuously in time. A third symmetry yields heteroclinic networks that consist of sets of all unstable attractors with that symmetry and the connections between them. Our results indicate that pulse-coupled systems can reliably generate well-defined sets of complex spatiotemporal patterns that conform to specific transition rules. We briefly discuss possible implications for computation with spiking neural systems.
Complex networks of dynamically connected saddle states persistently emerge in a broad range of high-dimensional systems and may reliably encode inputs as specific switching trajectories. Their computational capabilities, however, are far from being understood. Here, we analyze how symmetry-breaking inhomogeneities naturally induce predictable persistent switching dynamics across such networks. We show that such systems are capable of computing arbitrary logic operations by entering into switching sequences in a controlled way. This dynamics thus offers a highly flexible new kind of computation based on switching along complex networks of states.
Heteroclinic computing offers a novel paradigm for universal computation by collective system dynamics. In such a paradigm, input signals are encoded as complex periodic orbits approaching specific sequences of saddle states. Without inputs, the relevant states together with the heteroclinic connections between them form a network of states-the heteroclinic network. Systems of pulse-coupled oscillators or spiking neurons naturally exhibit such heteroclinic networks of saddles, thereby providing a substrate for general analog computations. Several challenges need to be resolved before it becomes possible to effectively realize heteroclinic computing in hardware. The time scales on which computations are performed crucially depend on the switching times between saddles, which in turn are jointly controlled by the system's intrinsic dynamics and the level of external and measurement noise. The nonlinear dynamics of pulse-coupled systems often strongly deviate from that of time-continuously coupled (e.g., phase-coupled) systems. The factors impacting switching times in pulse-coupled systems are still not well understood. Here we systematically investigate switching times in dependence of the levels of noise and intrinsic dissipation in the system. We specifically reveal how local responses to pulses coact with external noise. Our findings confirm that, like in time-continuous phase-coupled systems, piecewise-continuous pulse-coupled systems exhibit switching times that transiently increase exponentially with the number of switches up to some order of magnitude set by the noise level. Complementarily, we show that switching times may constitute a good predictor for the computation reliability, indicating how often an input signal must be reiterated. By characterizing switching times between two saddles in conjunction with the reliability of a computation, our results provide a first step beyond the coding of input signal identities toward a complementary coding for the intensity of those signals. The results offer insights on how future heteroclinic computing systems may operate under natural, and thus noisy, conditions.
The computation of rank ordering plays a fundamental role in cognitive tasks and offers a basic building block for computing arbitrary digital functions. Spiking neural networks have been demonstrated to be capable of identifying the largest k out of N analog input signals through their collective nonlinear dynamics. By finding partial rank orderings, they perform k-winners-take-all computations. Yet, for any given study so far, the value of k is fixed, often to k equal one. Here we present a concept for spiking neural networks that are capable of (re)configurable computation by choosing k via one global system parameter. The spiking network acts via pulse-suppression induced by inhibitory pulse-couplings. Couplings are proportional to each units' state variable (neuron voltage), constituting an uncommon but straightforward type of leaky integrate-and-fire neural network. The result of a computation is encoded as a stable periodic orbit with k units spiking at some frequency and others at lower frequency or not at all. Orbit stability makes the resulting analog-to-digital computation robust to sufficiently small variations of both, parameters and signals. Moreover, the computation is completed quickly within a few spike emissions per neuron. These results indicate how reconfigurable k-winners-take-all computations may be implemented and effectively exploited in simple hardware relying only on basic dynamical units and spike interactions resembling simple current leakages to a common ground. INDEX TERMS Analog computing, coupled oscillators, network dynamics, nonlinear dynamics, spiking neural networks, winner-takes-all, heteroclinic dynamics
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.