arXiv subjects

Yuansheng Lin

Publications and source records attributed to Yuansheng Lin.

3 recordsLinked to original sources

arXiv2017

Design of endurable networks in the presence of aging

Networks are designed to satisfy given objectives under specific requirements. While the static connectivity of networks is normally analyzed and corresponding design principles for static robustness are proposed, the challenge still remains of how to design endurable networks that maintain the required level of connectivity during its whole lifespan, against component aging. We introduce network endurance as a new concept to evaluate networks overall performance during its whole lifespan, considering both network connectivity and network duration. We develop a framework for designing an endurable network by allocating the expected lifetimes of its components, given a limited budget. Based on percolation theory and simulation, we find that the maximal network endurance can be achieved with a quantitative balance between network duration and connectivity. For different endurance requirements, we find that the optimal design can be separated into two categories: strong dependence of lifetime on node's degree leads to larger network lifetime, while weak dependence generates stronger network connectivity. Our findings could help network design, by providing a quantitative prediction of network endurance based on network topology.

physics.soc-ph↗

arXiv2017

Self-Organization of Dragon Kings

The mechanisms underlying cascading failures are often modeled via the paradigm of self-organized criticality. Here we introduce a simple model where nodes self-organize to be either weak or strong to failure which captures the trade-off between degradation and reinforcement of nodes inherent in many network systems. If strong nodes cannot fail, this leads to power law distributions of failure sizes with so-called "Black Swan" rare events. In contrast, if strong nodes fail once a sufficient fraction of their neighbors fail, this leads to "Dragon Kings", which are massive failures caused by mechanisms distinct from smaller failures. In our model, we find that once an initial failure size is above a critical value, the Dragon King mechanism kicks in, leading to piggybacking system-wide failures. We demonstrate that the size of the initial failed weak cluster predicts the likelihood of a Dragon King event with high accuracy and we develop a simple control strategy which also reveals that a random upgrade can inadvertently make the system more vulnerable. The Dragon Kings observed are self-organized, existing throughout the parameter regime.

physics.soc-ph↗

arXiv2014

Robustness of networks with topologies of dependency links

The robustness of complex networks with dependencies has been studied in recent years. However, previous studies focused on the robustness of networks composed of dependency links without network topology. In this study, we will analyze the percolation properties of a realistic network model where dependency links follow certain network topology. We perform the theoretical analysis and numerical simulations to show the critical effects of topology of dependency links on robustness of complex networks. For Erdös-Rényi (ER) connectivity network, we find that the system with dependency of RR topology is more vulnerable than system with dependency of ER topology. And RR-RR (i.e. random-regular (RR) network with dependency of RR topology) disintegrates in an abrupt transition. In particular, we find that the system of RR-ER shows different types of phase transitions. For system of different combinations, the type of percolation depends on the interaction between connectivity topology and dependency topology.

physics.soc-ph↗