Pei Wen-Jiang

A multiple objective optimization based echo state network tree and application to intrusion detection

Echo state network tree (ESNTree) is proposed in this paper. ESNTree changes the activation function of the hidden layer and modifies the initialization method of echo state network (ESN). If the number of input features is too large, genetic algorithm is used to extract better features. Thus the complexity of input feature space can be reduced. On the other hand, a divide-and-conquer method is used. Decision tree and ESN are combined to decrease the complexity of the classifier and make the classifier more comprehensible and more interpretable. Experiments show that ESNTree achieves better performance than neural network tree (NNTree). ESNTree has been applied to intrusion detection successfully.

A self-organizing shortest path finding strategy on complex networks

The shortcomings of traditional methods to find the shortest path are revealed, and a strategy of finding the self-organizing shortest path based on thermal flux diffusion on complex networks is presented. In our method, the shortest paths between the source node and the other nodes are found to be self-organized by comparing node temperatures. The computation complexity of the method scales linearly with the number of edges on underlying networks. The effects of the method on several networks, including a regular network proposed by Ravasz and Barabasi which is called the RB network, a real network, a random network proposed by Ravasz and Barabasi which is called the ER network and a scale-free network, are also demonstrated. Analytic and simulation results show that the method has a higher accuracy and lower computational complexity than the conventional methods.

Topological Self-Similar Networks Introduced by Diffusion-Limited Aggregation Mechanism

We propose a model for growing fractal networks based on the mechanisms learned from the diffusion-limited aggregation (DLA) model in fractal geometries in the viewpoint of network. By studying the DLA network, our model introduces multiplicative growth, aging and geographical preferential attachment mechanisms, whereby featuring topological self-similar property and hierarchical modularity. According to the results of theoretical analysis and simulation, the degree distribution of the proposed model shows a mixed degree distribution (i.e., exponential and algebraic degree distribution) and the fractal dimension and clustering coefficient can be tuned by changing the values of parameters.

Multiscale Entropy under the Inverse Gaussian Distribution: Analytical Results

The multiscale entropy (MSE) reveals the intrinsic multiple scales in the complexity of physical and physiological signals, which are usually featured by heavy-tailed distributions. However, most research results are pure experimental search. Recently, Costa et al. have made the first attempt to present the theoretical basis of MSE, but it only supports the Gaussian distribution [Phys Rev. E 71 (2005) 021906]. We present the theoretical basis of MSE under the inverse Gaussian distribution, a typical model for physiological, physical and financial data sets. The analysis allows for uncorrelated inverse Gaussian process and 1/f noise with the multivariate inverse Gaussian distribution, and then provides a reliable foundation for the potential applications of MSE to explore complex physical and physical time series.