Xu-Hua Yang

A NOVEL SMALL-WORLD NETWORK MODEL: KEEPING CONNECTIVITY WITHOUT ADDING EDGES

Considering the problems of potentially generating a disconnected network in the WS small-world network model [Watts and Strogatz, Nature393, 440 (1998)] and of adding edges in the NW small-world network model [Newman and Watts, Phys. Lett. A263, 341 (1999)], we propose a novel small-world network model. First, generate a regular ring lattice of N vertices. Second, randomly rewire each edge of the lattice with probability p. During the random rewiring procedure, keep the edges between the two nearest neighbor vertices, namely, always keep a connected ring. This model need not add edges and can maintain connectivity of the network at all times in the random rewiring procedure. Simulation results show that the novel model has the typical small-world properties which are small characteristic path length and high clustering coefficient. For large N, the model is approximately equal to the WS model. For large N and small p, the model is approximately equal to the WS model or the NW model.

A low complexity method for the optimization of network path length in spatially embedded networks

The average path length of a network is an important index reflecting the network transmission efficiency. In this paper, we propose a new method of decreasing the average path length by adding edges. A new indicator is presented, incorporating traffic flow demand, to assess the decrease in the average path length when a new edge is added during the optimization process. With the help of the indicator, edges are selected and added into the network one by one. The new method has a relatively small time computational complexity in comparison with some traditional methods. In numerical simulations, the new method is applied to some synthetic spatially embedded networks. The result shows that the method can perform competitively in decreasing the average path length. Then, as an example of an application of this new method, it is applied to the road network of Hangzhou, China.

Sequential support vector machine control of nonlinear systems by state feedback

Support vector machine is a new and promising technique for pattern classification and regression estimation. The training of support vector machine is characterized by a convex optimization problem, which involves the determination of a few additional tuning parameters. Moreover, the model complexity follows from that of this convex optimization problem. In this paper we introduce the sequential support vector machine for the regression estimation. The support vector machine is trained by the Kalman filter and particle filter respectively and then we design a controller based on the sequential support vector machine. Support vector machine controller is designed in the state feedback control of nonaffine nonlinear systems. The results of simulation demonstrate that the sequential training algorithms of support vector machine are effective and sequential support vector machine controller can achieve a satisfactory performance.

A visual automatic incident detection method on freeway based on RBF and SOFM neural networks

This paper proposes a novel visual automatic incident detection method on freeway based on RBF and SOFM neural networks. Two stages are involved. First, get the freeway traffic flow model based on the RBF neural networks and use the model to obtain the output prediction. The residuals will be gotten from the comparison between the actual and prediction. Second, use a SOFM neural networks to classify the residuals to detect the incident. Because the SOFM has the character of topological ordering, the winning neurons running trajectory on SOFM neuron array corresponds to the actual traffic state on freeway. We can observe the trajectory to detect the incident and achieve the visual traffic incident detection.

An ART2/RBF hybrid neural networks research

The radial basis function (RBF) neural networks have been widely used for approximation and learning due to its structural simplicity. However, there exist two difficulties in using traditional RBF networks: How to select the optimal number of intermediate layer nodes and centers of these nodes? This paper proposes a novel ART2/RBF hybrid neural networks to solve the two problems. Using the ART2 neural networks to select the optimal number of intermediate layer nodes and centers of these nodes at the same time and further get the RBF network model. Comparing with the traditional RBF networks, the ART2/RBF networks have the optimal number of intermediate layer nodes , optimal centers of these nodes and less error.