DeepRank: improving unsupervised node ranking via link discovery

A Node Selecting Approach for Traffic Network Based on Artificial Slime Mold

“…For a traffic network, node importance is connected with its propagation capability under all kinds of disturbances. Reference [4] put forward a DeepRank to improve unsupervised node ranking via link discovery. Hence, to evaluate the propagation capability of a traffic node may help us analyze the spreading dynamics and reconfigurable topology.…”

“…To rank the node importance, there are many traditional measures, including degree [1], location-aware [2], node content [3], link discovery [4], node centrality [6], topology-aware [7] and many other topology-based methods [8]. For example, [8] showed the impact of rank attack on network topology of routing protocol for low-power and lossy networks.…”

“…Reference [16] proposed a GEVD-based low-rank approximation for distributed adaptive node-specific signal estimation in wireless sensor networks, and [17] introduced FRANK as a fast node ranking approach in large-scale networks. Some researchers found it is important to collect the data on the traffic network and adopt a learning mechanism to realize topology development [3], [4]. [18] introduced an efficient mapping algorithm with a novel node-ranking approach for embedding virtual networks, and [19] used a spectral learning algorithm to reveal the propagation capability of complex networks.…”

“…But it is more difficult to efficiently evaluate the propagation capabilities by network data, i.e., [22] described an automated optimization of intersections using a genetic algorithm. The researches above applied a lot of complex artificial intelligence algorithms, such as deep learning [3], [4], spectral learning [19], genetic algorithm(GA) [22]. In 2010, a novel slime mold method was used for traffic network optimization [9]- [11].…”

“…Different from most living things using a brain or neurons to learn about the environment, the slime mold can only learn by a single-cell structure without any brain or neurons. The above learning algorithms [3], [4], [19], [22] are different from the single-cell slime mold, and cannot illustrate its operation mechanism. More importantly, the described slime mold in our paper is an artificial intelligence algorithm to solve the planning problems of traffic networks by simulating the foraging behavior of a real slime mold, rather than a true slime mold in [9]- [11], [23]- [25].…”