Applied Mechanics and Materials Vols. 411-414

Abstract: This paper proposed an adaptive differential evolution algorithm. The algorithm has an adaptive mutation factor which can be nonlinear reduced along with evolution process. Mutation factor is declined slowly in the beginning of evolution process in order to improve the global searching ability of the algorithm, and declined rapidly in the later of evolution process. The proposed algorithm is applied to solve flow shop scheduling to minimize makespan, computational experiments on a typical scheduling benchmark shows that the algorithm has a good performance.

Abstract: Recently, various papers investigated the topology identification and parameter identification of uncertain general complex dynamical networks. However, in many real complex dynamical network systems, there exists community or hierarchical structure and node delay. Based on LaSalle’s invariance principle, in this letter, an adaptive controlling method is proposed to identify unknown topological structure for general weighted complex dynamical network with community and node delay. Illustrative simulations are provided to verify the correctness and effectiveness of the proposed scheme.

Abstract: To solve prediction of sunspot number, a parallel process neural networks model is proposed in this paper, Firstly, by dividing the whole time-varying process into several small time intervals, the process neural networks are constructed in these small time intervals, which may disperse the load of networks. Then, employing the orthogonal basis expansion in functional space, the learning algorithm of the above-mentioned model is designed. The experimental results of time series predication of sunspots show that the proposed method has great potential for complicated nonlinear time series prediction.

Abstract: To enhance the approximation ability of process neural networks, a novel training algorithm is proposed by employing an improved quantum genetic algorithm. The proposed approach is applied to the training of process neural networks. The number of genes in a single chromosome is equal to the number of weight parameters. Taking each qubit in the current optimal chromosome as the goal, all individuals are updated by quantum rotation gate. In this method, each chromosome has three chains of genes, which can accelerate convergence. Taking the pattern classification of trigonometric functions as an example, the experimental results show that the proposed method is obviously superior to the common process neural networks.

Abstract: A hidden Markov model (HMM) encompasses a large class of stochastic process models and has been successfully applied to a number of scientific and engineering problems, including speech and other pattern recognition problems, and biological sequence analysis. A major restriction is found, however, in conventional HMM, i.e., it is ill-suited to capture the interactions among different models. A variety of coupled hidden Markov models (CHMMs) have recently been proposed as extensions of HMM to better characterize multiple interdependent sequences. The resulting models have multiple state variables that are temporally coupled via matrices of conditional probabilities. This paper study is focused on the coupled discrete HMM, there are two state variables in the network. By generalizing forward-backward algorithm, Viterbi algorithm and Baum-Welch algorithm commonly used in conventional HMM to accommodate two state variables, several new formulae solving the 2-chain coupled discrete HMM probability evaluation, decoding and training problem are theoretically derived.

Abstract: The Theory of Projection Pursuit Regression is applied in the equipment indemnificatory valuation and forecast to establish the projection pursuit regression model. After fitting the training samples, this model strikes a good balance between the valuation value and its relevant influential factors, demonstrating a good fitting effect with the average relative error of only 2.1522% . After predicting the test samples, it shows a good forecast effect with the relative error of only-0.4069%, thus providing basis for equipment indemnificatory valuation and forecast.

Abstract: In this paper, we propose a unified approach for computing the t/t-diagnosability of numerous multiprocessor systems under the PMC model, including hypercube-like graphs, star graphs, and pancake graphs. Our approach first defines a superclass of graphs, called j-order cluster-composition graphs, to cover them.We then show that the 1-order simple cluster-composition graph is t/t-diagnosable if it contains no connected component with size less than 2t+1, where t is the minimal number of neighbors of any pair of vertices of the graph. Based on this result, the t/t-diagnosability of the above multiprocessor systems can be computed efficiently.

Abstract: Plan recognition is an important problem to address in order to enhance the capabilities of battlefield surveillance systems. It is concerned with finding a priori defined plan that possibly are instantiated in the present flow of battle events. Plan recognition requires that many sub problems are solved. For instance, we need to establish which situations are interesting, how to represent these situations, and inferable events and states that can be used for representing them. In this paper we discuss current research efforts and goals concerning template-based plan recognition. We provide a categorization of approaches for plan recognition together with a formalization of the template-based plan recognition problem. We discuss this formalization in the light of an air-ground striking scenario. Finally, we conclude that plan recognition is an important problem to look into for enhancing the overall situation awareness of decision makers.

Abstract: A synthetical efficient mapping from individual preferences to group preference isconstructed in this paper by means of group synthetical efficient number of alternatives.Based onsome fundamental properties of mapping ,a method of synthetical efficient ordering alternative for group multiobjective decision making in given .

Abstract: This paper introduces a method which uses the gene expression programming algorithm to conduct multivariate nonlinear function modeling, which is applied in the earthquake magnitude prediction. The experiment shows that the prediction accuracy of the GEP is significantly higher than that of the neural network model. Finally, by using the non-delayed effects and stability of the earthquake magnitude prediction data, the state-transition matrix is obtained through the Markov chain, and the state interval and corresponding probability of the GEP model prediction are obtained. In this way, the credibility of the prediction results has been increased.