Neural Computing and Applications

Li, Shuai; Liu, Bo; Chen, Baogang; Lou, Yuesheng

doi: 10.1007/s00521-012-0950-1pmid: N/A

Location information is useful for mobile phones. There exists a dilemma between the relatively high price of GPS devices and the dependence of location information acquisition on GPS for most phones in current stage. To tackle this problem, in this paper, we investigate the position inference of phones without GPS according to Bluetooth connectivity and positions of beacon phones. With the position of GPS-equipped phones as beacons and with the Bluetooth connections between neighbor phones as constraints, we formulate the problem as an optimization problem defined on the Bluetooth network. The solution to this optimization problem is not unique. Heuristic information is employed to improve the performance of the result in the feasible set. Recurrent neural networks are developed to solve the problem distributively in real time. The convergence of the neural network and the solution feasibility to the defined problem are both theoretically proven. The hardware implementation of the proposed neural network is also explored in this paper. Simulations and comparisons with different application backgrounds are considered. The results demonstrate the effectiveness of the proposed method.

Khan, Yasir; Vázquez-Leal, Héctor; Wu, Q.

doi: 10.1007/s00521-012-0952-zpmid: N/A

The purpose of this study is to introduce an efficient iterated homotopy perturbation transform method (IHPTM) for solving a mathematical model of HIV infection of CD4+ T cells. The equations are Laplace transformed, and the nonlinear terms are represented by He’s polynomials. The solutions are obtained in the form of rapidly convergent series with elegantly computable terms. This approach, in contrast to classical perturbation techniques, is valid even for systems without any small/large parameters and therefore can be applied more widely than traditional perturbation techniques, especially when there do not exist any small/large quantities. A good agreement of the novel method solution with the existing solutions is presented graphically and in tabulated forms to study the efficiency and accuracy of IHPTM. This study demonstrates the general validity and the great potential of the IHPTM for solving strongly nonlinear problems.

Zhao, Mo; Zhang, Huaguang; Wang, Zhiliang

doi: 10.1007/s00521-012-0964-8pmid: N/A

This paper proposes a new approach of synchronization in complex dynamical networks. In this method, the scalar signals are used to instead the output variables of every node as the feedback variables and transmitted signals between every two coupling nodes. As a result, it not only simplifies the topological structure but also saves channel resources at the same time. Especially, some of the criteria are expressed in normal algebraic inequalities instead of matrix inequalities, which means that the original computational effort required is greatly decreased. Finally, several simulation examples are provided to show the effectiveness of the proposed results.

Yin, Fei; Jiao, L.; Shang, Fanhua; Wang, Shuang; Hou, Biao

doi: 10.1007/s00521-012-0978-2pmid: N/A

Recently, there has been a lot of interest in the underlying sparse representation structure in high-dimensional data such as face images. In this paper, we propose two novel efficient dimensionality reduction methods named Fast Sparsity Preserving Projections (FSPP) and Fast Fisher Sparsity Preserving Projections (FFSPP), respectively, which aim to preserve the sparse representation structure in high-dimensional data. Unlike the existing Sparsity Preserving Projections (SPP), where the sparse representation structure is learned through resolving n (the number of samples) time-consuming $$ \ell^{ 1} $$ norm optimization problems, FSPP constructs a dictionary through classwise PCA decompositions and learns the sparse representation structure under the constructed dictionary through matrix–vector multiplications, which is much more computationally tractable. FFSPP takes into consideration both the sparse representation structure and the discriminating efficiency by adding the Fisher constraint to the FSPP formulation to improve FSPP’s discriminating ability. Both of the proposed methods can boil down to a generalized eigenvalue problem. Experimental results on three publicly available face data sets (Yale, Extended Yale B and ORL), and a standard document collection (Reuters-21578) validate the feasibility and effectiveness of the proposed methods.

Tavoosi, Jafar; Badamchizadeh, Mohammad

doi: 10.1007/s00521-012-0981-7pmid: N/A

This paper presents the ability of the interval type-2 Takagi–Sugeno–Kang fuzzy neural networks (IT2-TSK-FNN) for nonlinear dynamical system identification. The proposed IT2-TSK-FNN has seven layers. The first two layers consist of type-2 fuzzy neurons with uncertainty in the mean of Gaussian membership functions. Third layer is rule layer. Type-reduction is done in fourth layer. In the fifth, sixth, and seventh layers, consequent left–right firing points, two end points, and output are evaluated, respectively. In this paper, gradient descent with adaptive learning rate backpropagation is used in learning phase. IT2-TSK-FNN is used for the identification of three nonlinear systems, and then results are compared with adaptive-network-based fuzzy inference system (ANFIS).

Uzer, Mustafa; Inan, Onur; Yılmaz, Nihat

doi: 10.1007/s00521-012-0982-6pmid: N/A

Two hybrid feature selection methods (SFSP and SBSP) which are composed by combining the sequential forward selection and the sequential backward selection together with the principal component analysis developed by utilizing quadratic discriminant analysis classification algorithmic criteria so as to utilize in the diagnosis of breast cancer fast and effectively are presented in this study. The tenfold cross-validation method has been applied in the algorithm, which is utilized as criteria during the selection of the features. The dimension of the feature space for input has been decreased from 9 to 4 thanks to the selection of these two hybrid features. The Artificial Neural Networks have been used as classifier. The cross-validation method has been preferred also in the phase of this classification as in the case of the selection of the feature in order to increase the reliability of the result. The Wisconsin Breast Cancer Database obtained from the UCI has been utilized so as to determine the correctness of the system suggested. The values of the average correctness of the classification obtained by utilizing a tenfold cross-validation of the two hybrid systems developed earlier are found, respectively, as follows: for SFSP + NN, 97.57 % and for SBSP + NN, 98.57 %. SBSP + NN system has been observed that, among the studies carried out by implementing the cross-validation method for the breast cancer, the result appears to be very promising. The acquired results have revealed that this hybrid system applied by means of reducing dimension is an utilizable system in order to diagnose the diseases faster and more successfully.

Chen, Yan-Cheng; Su, Chao-Ton; Yang, Taho

doi: 10.1007/s00521-012-0985-3pmid: N/A

Support vector machines (SVMs) are state-of-the-art tools used to address issues pertinent to classification. However, the explanation capabilities of SVMs are also their main weakness, which is why SVMs are typically regarded as incomprehensible black box models. In the present study, a rule extraction algorithm to extract the comprehensible rule from SVMs and enhance their explanation capability is proposed. The proposed algorithm seeks to use the support vectors from a training model of SVMs and combine genetic algorithms for constructing rule sets. The proposed method can not only generate rule sets from SVMs based on the mixed discrete and continuous variables but can also select important variables in the rule set simultaneously. Measurements of accuracy, sensitivity, specificity, and fidelity are utilized to compare the performance of the proposed method with direct learner algorithms and several rule-extraction techniques from SVMs. The results indicate that the proposed method performs at least as well as with the most successful direct rule learners. Finally, an actual case of pressure ulcer was studied, and the results indicated the practicality of our proposed method in real applications.

Tsai, Hsing-Chih; Wu, Yun-Wu; Tyan, Yaw-Yauan; Lin, Yong-Huang

doi: 10.1007/s00521-012-0987-1pmid: N/A

This study designed a four-layer modular neural network (MNN) to predict and program squat wall strength values. Results generated by the proposed MNN include predictions and programmed formulas that are similar in form to modular polynomials, which permit MNN programming to interpret training results in a meaningful way that offers significant advantages over famous neural networks. This study employed particle swarm optimization for MNN parameter learning and structure learning in order to prune MNN to avoid overfitting and increase programmed formula concision. To extend the uses of MNN programming, this paper further employed MNN tuning to refine existing analytical methods and codes. Case studies focused on squat wall strength analyses. Study results demonstrated that MNN programming uniquely uses a programmed formula to deliver good prediction accuracy. MNN tuning further improved the studied methods. Programmed formulas also provided insights into input parameter impacts and significant modular functions.

Chang, Jhih-Chung

doi: 10.1007/s00521-012-0988-0pmid: N/A

The computational complexity of the optimum maximum-likelihood detector does not allow its utility for multiuser detection (MUD) in code-division multiple-access (CDMA) systems. In this paper, a novel particle swarm optimization (PSO) algorithm is suggested to carry out MUD for synchronous CDMA systems. This work considers two new aspects, namely an adaptive velocity updating mechanism based on Newton method and a dynamic inertia weight into the standard PSO. These mechanisms can provide more diversity to help avoiding premature convergence and significantly improve the bit error rate performance. Several computer simulation results demonstrate that the proposed modified PSO detector significantly outperforms the decorrelating detector, the linear minimum mean square error detector, and the standard PSO-based detector.

Miao, Guoying; Wang, Zhen; Ma, Qian; Lu, Junwei

doi: 10.1007/s00521-012-0991-5pmid: N/A

In this paper, consensus problems of multi-agent systems with nonlinear dynamics and time delays are investigated. Two cases are considered that are the system without a leader and the one with a leader. New linear control protocols are introduced to make second-order multi-agent systems consensus. By applying the Lyapunov function method and the graph theory, some sufficient conditions for consensus of positions and velocities are derived. Finally, two simulation results are given to illustrate the effectiveness of our proposed methods.