Small-world networks exhibit pronounced intermittent synchronization

Abnormal suppression of thermal transport by long-range interactions in networks

Heat and electricity are two fundamental forms of energy widely utilized in our daily lives. Recently, in the study of complex networks, there is growing evidence that they behave significantly different at the micro-nanoscale. Here, we use a small-world network model to investigate the effects of reconnection probability p and decay exponent α on thermal and electrical transport within the network. Our results demonstrate that the electrical transport efficiency increases by nearly one order of magnitude, while the thermal transport efficiency falls off a cliff by three to four orders of magnitude, breaking the traditional rule that shortcuts enhance energy transport in small-world networks. Furthermore, we elucidate that phonon localization is a crucial factor in the weakening of thermal transport efficiency in small-world networks by characterizing the density of states, phonon participation ratio, and nearest-neighbor spacing distribution. These insights will pave new ways for designing thermoelectric materials with high electrical conductance and low thermal conductance.

Criticality in reservoir computer of coupled phase oscillators

Accumulating evidence shows that the cerebral cortex is operating near a critical state featured by power-law size distribution of neural avalanche activities, yet evidence of this critical state in artificial neural networks mimicking the cerebral cortex is still lacking. Here we design an artificial neural network of coupled phase oscillators and, by the technique of reservoir computing in machine learning, train it for predicting chaos. It is found that when the machine is properly trained, oscillators in the reservoir are synchronized into clusters whose sizes follow a power-law distribution. This feature, however, is absent when the machine is poorly trained. Additionally, it is found that despite the synchronization degree of the original network, once properly trained, the reservoir network is always developed to the same critical state, exemplifying the "attractor" nature of this state in machine learning. The generality of the results is verified in different reservoir models and by different target systems, and it is found that the scaling exponent of the distribution is independent of the reservoir details and the bifurcation parameters of the target system, but is modified when the dynamics of the target system is changed to a different type. The findings shed light on the nature of machine learning, and are helpful to the design of high-performance machines in physical systems.

Complex networks exhibit intermittent synchronization

The path toward the synchronization of an ensemble of dynamical units goes through a series of transitions determined by the dynamics and the structure of the connections network. In some systems on the verge of complete synchronization, intermittent synchronization, a time-dependent state where full synchronization alternates with non-synchronized periods, has been observed. This phenomenon has been recently considered to have functional relevance in neuronal ensembles and other networked biological systems close to criticality. We characterize the intermittent state as a function of the network topology to show that the different structures can encourage or inhibit the appearance of early signs of intermittency. In particular, we study the local intermittency and show how the nodes incorporate to intermittency in hierarchical order, which can provide information about the node topological role even when the structure is unknown.

Echo in complex networks

Large populations of globally coupled or uncoupled oscillators have been recently shown to exhibit an intriguing echo behavior [Ott, Platig, Antonsen, and Girvan, Chaos: An Interdiscip. J. Nonlinear Sci. 18, 037115 (2008)CHAOEH1054-150010.1063/1.2973816; Chen, Tinsley, Ott, and Showalter, Phys. Rev. X 6, 041054 (2016)2160-330810.1103/PhysRevX.6.041054], wherein a system is perturbed by two successive pulses at times T and T+τ inducing a spontaneous increase in the order parameter at the given times. These two provoked increments in the order parameter are followed by an unprovoked spontaneous increment in the order parameter at time T+2τ termed as an echo. In this paper, the effects of network topology on the emergence of an echo are explored. Two principal network parameters, namely, average degree and network randomness, are varied for this purpose. The networks are rewired to increase randomness in the network connections using the Watts-Strogatz algorithm to generate small world networks [Watts and Strogatz, Nature (London) 393, 440 (1998)10.1038/30918]. Thus, the whole span of networks ranging from a regular ring to a completely random network is explored. The average degree of the underlying connectivity, starting from nearest neighbor connections, is also monotonically increased and its effects on the echo behavior are analyzed. We find that for rings with low average degrees and high coupling strengths a discernible echo is not observed. Remarkably, an echo reemerges in the presence of sufficient randomness in the connections for such networks. For a regular ring network, increasing the average degree after a critical value also yields a transition to echo behavior. However, for completely random networks echoes are present in networks of all average degrees. This suggests that randomizing connections can induce echoes in systems even when the average degree of connections is very low. Another subtle feature arises for intermediate randomness, where the system exhibits a nonmonotonic dependence of the echo size on average degree. The echo size was found to be minimum at an intermediate value of the average degree. Lastly we consider the influence of dynamically changing links on the echo size and demonstrate that time-varying connections destroy the echo in low average degree networks, while the echo persists under dynamic links in high average degree networks. So our results clearly demarcate the class of networks that are robust candidates for exhibiting echoes, as well as provide caveats for the observation of echoes in networks.