Search

An optimization algorithm for dynamic Bayesian networks (DBN) based on Bayesian optimization algorithm (BOA) is developed for learning and constructing the DBN structure. In this paper, we first introduce some basic theories and concepts of probability model evolutionary algorithm. Then we describe, the basic mode for constructing DBN diagram and the mechanism of DBN structure learning based on BOA. The DBN structure learning based on BOA consists of two parts. The first part is to obtain the structure and parameters of DBN in terms of a good solution, and the second part is to produce new groups according to the obtained DBN structure. In this paper, the DBN learning is achieved by genetics algorithm based on a greedy mechanism. The DBN inference is performed by a forward-simulation algorithm. Simulation results are provided to demonstrate the effectiveness of the proposed algorithm. [ABSTRACT FROM AUTHOR]

Generalization problem is a key problem in NN society, which can be grouped into two classes: the generalization problem with unlimited size of training sample and that with limited size of training sample. The generalization problem with limited size of training sample is considered in this paper. Similar to margin maximization criterion in SVM, we propose a margin maximization training algorithm for BP network to further improve the generalization ability of BP network. Experimental results show that the margin maximization training algorithm proposed in this paper does improve the performance of BP network, and shows a comparable performance with SVM. [ABSTRACT FROM AUTHOR]

Chaotic noise cancellation has potential application in both secret communication and radar target identification. To solve the problem of parameter estimation in chaotic noise, a novel radial basis function neural network (RBF-NN) -based chaotic time series data modeling method is presented in this paper. Together with the spectral analysis technique, the algorithm combines neural network's ability to approximate any nonlinear function. Based on the flexibility of RBF-NN predictor and classical amplitude spectral analysis technique, this paper proposes a new algorithm for parameter estimation in chaotic noise. Analysis of the proposed algorithm's principle and simulation experiments results are given out, which show the effective of the proposed method. We conclude that the study has potential application in various fields as in secret communication for narrow band interference rejection or attenuation and in radar signal processing for weak target detection and identification in sea clutter. [ABSTRACT FROM AUTHOR]

Detecting the boundaries of protein domains has been an important and challenging problem in experimental and computational structural biology. In this paper the domain detection is first taken as an imbalanced data learning problem. A novel undersampling method using distance-based maximal entropy in the feature space of SVMs is proposed. On multiple sequence alignments that are derived from a database search, multiple measures are defined to quantify the domain information content of each position along the sequence. The overall accuracy is about 87% together with high sensitivity and specificity. Simulation results demonstrate that the utility of the method can help not only in predicting the complete 3D structure of a protein but also in the machine learning system on general imbalanced datasets. [ABSTRACT FROM AUTHOR]

Provide these intelligent medical services, it is necessary to understand the situation information generated in a hospital. There should be infra technologies that can classify and control the information for processing situation data, not mere collection of conceptual information, with clear standards. This paper, as a study to seize the information generated from medical situation more clearly, understood the property of data using neural network and applied the security ratings of information so that the system to provide the user appropriate to designated rating with analyzed medical information is established. It will be an effective measure to enhance the effectiveness of medical devices and backup data already introduced and understand the various medical data that will be generated from medical devices to be introduced. [ABSTRACT FROM AUTHOR]

An edge detection method by combining image entropy and Self -Organizing Map (SOM) is proposed in this paper. First, according to information theory image entropy is used to curve up the smooth region and the region of gray level abruptly changed. Then we transform the gray level image to ideal binary pattern of pixels. We define six classes' edge and six edge prototype vectors. These edge prototype vectors are fed into input layer of the Self-Organizing Map (SOM). Classifying the type of edge through this network, the edge image is obtained. At last, the speckle edges are discarded from the edge image. Experimental results show that it gained better edge image compared with Canny edge detection method. [ABSTRACT FROM AUTHOR]

With the development of Intelligent Transport Systems (ITS), automatic license plate recognition (LPR) plays an important role in numerous applications in reality. In this paper, a coarse to fine algorithm to detect license plates in images and video frames with complex background is proposed. First, the method based on Component Connect (CC) is used to detect the possible license plate regions in the coarse detection. Second, the method based on texture analysis is applied in the fine detection. Finally, a BP Neural Network is adopted as classifier, parts of the features is selected based on statistic diagram to make the network efficient. The average accuracy of detection is 95.3% from the images with different angles and different lighting conditions. [ABSTRACT FROM AUTHOR]

Of the possible interactions between human and robot, touch is an important means of providing human beings with emotional relief. However, most previous studies have focused on interactions based on voice and images. In this paper, a method of recognizing human touching behaviors is proposed for developing a robot that can naturally interact with humans through touch. In this method, the recognition process is divided into pre-process phase and recognition phase. In the pre-process phase, recognizable characteristics are calculated from the data generated by the touch detector which was fabricated using force sensors. The force sensor used an FSR (force sensing register). The recognition phase classifies human touching behaviors using a multi-layer perceptron which is a neural network model. We measured three different human touching behaviors for six men. The human touching behaviors are ‘hitting,' 'stroking,' and ‘tickling'. In the test conducted with recognizers generated for each user, the average recognition rate was 93.8%, while the test conducted with a single recognizer showed a 79.8% average recognition rate. These results show the feasibility of the proposed human touching behavior recognition method. [ABSTRACT FROM AUTHOR]

This paper investigates the construction of a wide class of singlehidden layer neural networks (SLNNs) with or without tunable parameters in the hidden nodes. It is a challenging problem if both the parameter training and determination of network size are considered simultaneously. Two alternative network construction methods are considered in this paper. Firstly, the discrete construction of SLNNs is introduced. The main objective is to select a subset of hidden nodes from a pool of candidates with parameters fixed ‘a priori'. This is called discrete construction since there are no parameters in the hidden nodes that need to be trained. The second approach is called continuous construction as all the adjustable network parameters are trained on the whole parameter space along the network construction process. In the second approach, there is no need to generate a pool of candidates, and the network grows one by one with the adjustable parameters optimized. The main contribution of this paper is to show that the network construction can be done using the above two alternative approaches, and these two approaches can be integrated within a unified analytic framework, leading to potentially significantly improved model performance and/or computational efficiency. [ABSTRACT FROM AUTHOR]

At present, most of the algorithms for learning Bayesian Networks (BNs) use EM algorithm to deal with incomplete data. They are of low efficiency because EM algorithm has to perform iterative process of probability reasoning to complete the incomplete data. In this paper we present an efficient BN learning algorithm, which use the combination of EMI method and a scoring function based on mutual information theory. The algorithm first uses EMI method to estimate, from incomplete data, probability distributions over local structures of BNs, then evaluates BN structures with the scoring function and searches for the best one. The detailed procedure of the algorithm is depicted in the paper. The experimental results on Asia and Alarm networks show that when achieving high accuracy, the algorithm is much more efficient than two EM based algorithms, SEM and EM-EA algorithms. [ABSTRACT FROM AUTHOR]

Fuzzy Petri net (FPN) is a powerful modeling tool for fuzzy production rules based knowledge systems. But it is lack of learning mechanism, which is the main weakness while modeling uncertain knowledge systems. Fuzzy neural Petri net (FNPN) is proposed in this paper, in which fuzzy neuron components are introduced into FPN as a sub-net model of FNPN. For neuron components in FNPN, back propagation (BP) learning algorithm of neural network is introduced. And the parameters of fuzzy production rules in FNPN neurons can be learnt and trained by this means. At the same time, different neurons on different layers can be learnt and trained independently. The FNPN proposed in this paper is meaningful for Petri net models and fuzzy systems. [ABSTRACT FROM AUTHOR]

Consumer demand for intelligent battery charges is increasing as portable electronic applications continue to grow. Fast charging of battery packs is a problem which is difficult, and often expensive, to solve using conventional techniques. Conventional techniques only perform a linear approximation of a nonlinear behavior of a battery packs. The battery charging is a nonlinear electrochemical dynamic process and there is no exact mathematical model of battery. Better techniques are needed when a higher degree of accuracy and minimum charging time are desired. In this paper we propose soft computing approach based on fuzzy recurrent neural networks (RFNN) training by genetic algorithms to control batteries charging process. This technique does not require mathematical model of battery packs, which are often difficult, if not impossible, to obtain. Nonlinear and uncertain dynamics of the battery pack is modeled by recurrent fuzzy neural network. On base of this FRNN model, the fuzzy control rules of the control system for battery charging is generated. Computational experiments show that the suggested approach gives least charging time and least Tend-Tstart results according to the other intelligent battery charger works. [ABSTRACT FROM AUTHOR]

In this paper, synchronization is investigated for an array of nonlinearly coupled identical connected neural networks with delay. By employing the Lyapunov functional method and the Kronecker product technique, several sufficient conditions are derived. It is shown that global exponential synchronization of the coupled neural networks is guaranteed by a suitable design of the coupling matrix, the inner linking matrix and some free matrices representing the relationships between the system matrices. The conditions obtained in this paper are in the form of linear matrix inequalities, which can be easily computed and checked in practice. A typical example with chaotic nodes is finally given to illustrate the effectiveness of the proposed synchronization scheme. [ABSTRACT FROM AUTHOR]

Electroencephalogram (EEG) is a common tool to explore brain activities ranging from concentrated cognitive efforts to sleepiness. For the issue of sleepiness, pupil behavior can provide some information regarding alertness. The issue of sleepiness can be reflected by EEG energy. Specifically, intrusion of EEG theta wave activity into the beta activity of active wakefulness has been interpreted as ensuing sleepiness. This paper develops an innovative signal classification method that is capable of differentiating subjects with sleep disorder of obstructive sleep apnea (OSA) which cause excessive daytime sleepiness (EDS) from normal control subjects who do not have a sleep disorder. The theta energy ratios are calculated from the 2-second sliding windows by Fourier transform. An artificial neural network of modified ART2 is utilized to identify subjects with OSA from a combined group of subjects including healthy controls. This grouping from the neural network is then compared with the actual diagnostic classification of subjects as OSA or healthy controls and it is found to be 91% accurate in differentiating between the two groups. [ABSTRACT FROM AUTHOR]

Accurate classification of samples using gene expression frofiles is very important in cancer detection and treatment. In this paper, a novel EPA-KNN (Emerging Patterns Advanced-K Nearest Neighbors) gene classification algorithm is proposed. Bayes estimation is applied for the computation of entropy to improve its reliability, and RCP (Random Cut Point) is presented to strengthen the generalization about unknown test samples. With these improvements, the EPAs are acquired. Then an EPA based classifier is constructed inspired by KNN. The experimental results show the new algorithm is feasible and effective. [ABSTRACT FROM AUTHOR]

The detection of explosives and illicit material in passengers' luggage for the purpose of station security is an important area in public traffic security. This paper presents a method for X-ray image recognition based on fuzzy rule and parallel neural networks. Neural networks have been widely used in various fields. However, the computing efficiency decreases rapidly if the scale of neural network increases. In this paper, a new method of X-ray image recognition based on the fuzzy-neuron system is proposed. In fuzzy rules method, a test pattern may belong to several classes with different degrees. A neural networks classifier is just for one class and used to make sure if the pattern is really belonged to that class based on fuzzy rules, they are combined to obtain the recognition result. From the experience results, the new method performs well. [ABSTRACT FROM AUTHOR]

Micorarray data are often extremely asymmetric in dimensionality, such as thousands or even tens of thousands of genes and a few hundreds of samples. Such extreme asymmetry between the dimensionality of genes and samples presents several challenges to conventional clustering and classification methods. In this paper, a novel ensemble method based on correlation analysis is proposed. Firstly, in order to extract useful features and reduce dimensionality, different feature selection methods based on correlation analysis are used to form different feature subsets. Then a pool of candidate base classifiers is generated to learn the subsets which are re-sampling from the different feature subsets. At last, appropriate classifiers are selected to construct the classification committee using EDA (Estimation of Distribution Algorithms) algorithm. Experiments show that the proposed method produces the best recognition rates on two benchmark databases. [ABSTRACT FROM AUTHOR]

Air-pressure limit value is an important conditional parameter of artificial respiration. The pulmonary characteristics are very different according to the person. For setting appropriate ventilation conditions fitting to each patient, it is necessary to establish a mathematical model describing the mechanism of human respiratory system, and to know the pulmonary characteristic of each patient via identification of the model. For this purpose, two types of respiratory system models have been proposed by the authors. These models are expressed as second order nonlinear differential equations with air-volume variant elastic coefficient and air-volume variant resistive coefficient. In the first type of model, elastic coefficient is expressed as polynomial function of air-volume, while in the second type of model, elastic coefficient is expressed by RBF network. The model with polynomial expression of elastance has the advantage that the structure is simple. On the other hand, the model with RBF expression of elastance has better numerical stability against to the model with polynomial expression of elastance. In this paper, a decision method of air-pressure limit value based on the respiration model with RBF expression of elastance is proposed. This method adopt a numerical technique to find the point of saturation starting point in the elastance curve, So direct calculation of radius of curvature can be avoided. The proposed method is validated by an example of application to practical clinical data. [ABSTRACT FROM AUTHOR]

Many different approaches for the modeling of neural networks were presented in the literature (e.g. [4]). Generally the object-oriented approach proved itself as most appropriate. It provides a concise but comprehensive framework for the design of neural networks in terms of its static and dynamic components, i.e. the information structure and its methods in the object-oriented notion. This paper presents a framework for the conceptual and physical integration of neural networks into object-relational database systems. The static components comprise the structural parts of a neural network, as the neurons and connections, higher topological structures as layers, blocks and network systems. The dynamic components are the behavioral characteristics, as the creation, training and evaluation of the network. Finally the implementation of the new NeuroOracle system based on the proposed framework is presented. [ABSTRACT FROM AUTHOR]

In this paper we propose a comparative study of Artificial Neural Networks (ANN) and Artificial Immune Systems. Artificial Immune Systems (AIS) represent a novel paradigm in the field of computational intelligence based on the mechanisms that allow vertebrate immune systems to face attacks from foreign agents (called antigens). Several similarities as well as differences have been shown by Dasgupta in [1]. Here we present a comparative study of these two approaches considering evolutions of the concepts of ANN and AIS, respectively hybrid neural systems, Artificial Immune Recognition Systems (AIRS) and aiNet. We tried to establish a comparison among these three methods using a well known dataset, namely the Wisconsin Breast Cancer Database. We observed interesting trends in systems' performances and capabilities. Peculiarities of these systems have been analyzed, possible strength points and ideal contexts of application suggested. These and other considerations will be addressed in the rest of this manuscript. [ABSTRACT FROM AUTHOR]

With Kohonen's self-organizing map based brain MRI image segmentation, there are still some regions which are not partitioned accurately, particularly in the transitional regions of gray matter and white matter, or cerebrospinal fluid and gray matter. In this paper, we propose a dynamically growing hierarchical self-organizing map integrated with a multi-scale feature vector to overcome the problem mentioned above, which uses the spatial relationships between image pixels and using multi-scale processing method to reduce the noise effect and the classification ambiguity. The efficacy of our approach is validated by extensive experiments using both simulated and real MRI images. [ABSTRACT FROM AUTHOR]

The ability to predict traffic variables such as speed, travel time and flow, based on real time and historic data, collected by various systems in transportation networks, is vital to the intelligent transportation systems (ITS). The present paper proposes a method based on Principal Component Analysis and Support Vector Regression (PCA-SVR) for a short-term simultaneously prediction of network traffic flow which is multidimensional compared with traditional single point. Data from a typical traffic network of Beijing City, China are used for the analysis. Other models such as ANN and ARIMA are also developed as a comparison of the performance of both these techniques is carried out to show the effectiveness of the novel method. [ABSTRACT FROM AUTHOR]

Extracting proper features is crucial to the performance of a pattern recognition system. Since the goal of a pattern recognition system is to recognize a pattern correctly, a natural measure of "goodness" of extracted features is the probability of classification error. However, popular feature extraction techniques like principal component analysis (PCA), Fisher linear discriminant analysis (FLD), and independent component analysis (ICA) extract features that are not directly related to the classification accuracy. In this paper, we present two linear discriminant analysis algorithms (LDA) whose criterion functions are directly based on minimum probability of classification error, or the Bayes error. We term these two linear discriminants as recursive Bayesian linear discriminant I (RBLD-I) and recursive Bayesian linear discriminant II (RBLD-II). Experiments on databases from UCI Machine Learning Repository show that the two novel linear discriminants achieve superior classification performance over recursive FLD (RFLD). [ABSTRACT FROM AUTHOR]

In recent years, many efforts have been put in applying the concept of reconfigurable computing to neural networks. In our previous pursuits, an innovative self-organizing learning array (SOLAR) was developed. However, traditional multiplexer method to achieve reconfigurable connection has its limit for larger networks. In this paper, we propose a novel pipeline structure, which offers flexible, possibly large number of dynamically configurable connections and which utilizes each node's computing ability. The hardware resources demand of the proposed structure is a linear function of the network size, which is especially useful for building a large network that can handle complicated real-world applications. [ABSTRACT FROM AUTHOR]

Neighborhood Preserving Embedding (NPE) is a subspace learning algorithm. Since NPE is a linear approximation to Locally Linear Embedding (LLE) algorithm, it has good neighborhood-preserving properties. Although NPE has been applied in many fields, it has limitations to solve recognition task. In this paper, a novel subspace method, named Kernel Fisher Neighborhood Preserving Embedding (KFNPE), is proposed. In this method, discriminant information as well as the intrinsic geometry relations of the local neighborhoods are preserved according to prior class-label information. Moreover, complex nonlinear variations of real face images are represented by nonlinear kernel mapping. Experimental results on ORL face database demonstrate the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

Response Modeling is concerned with computing the likelihood of a customer to respond to a marketing campaign. A major problem encountered in response modeling is huge volume of data or patterns. The k-NN has been used in various classification problems for its simplicity and ease of implementation. However, it has not been applied to problems for which fast classification is needed since the classification time rapidly increases as the size of reference set increases. In this paper, we propose a clustering-based preprocessing step in order to reduce the size of reference set. The experimental results showed an 85% decrease in classification time without a loss of accuracy. [ABSTRACT FROM AUTHOR]

This paper proposes a novel recognition method for palmprints using a new sparse coding (SC) algorithm proposed by us. This algorithm exploited the maximum Kurtosis as the sparse measure criterion, at one time, a fixed variance term of sparse coefficients is used to yield a fixed information capacity. Experimental results show that the feature basis vectors of palmprint images can be successfully extracted by using our SC algorithm. Using the radial basis probabilistic neural network (RBPNN), the classification task can be implemented easily. Finally, compared with methods of principal component analysis (PCA) and the classical SC, simulation results show that our algorithm is indeed efficient and effective in performing palmprint recognition task. [ABSTRACT FROM AUTHOR]

Bayesian subspace analysis (BSA) has been successfully applied in data mining and pattern recognition. However, due to the use of probabilistic measure of similarity, it often needs much more projective vectors for better performance, which makes the compression ratio very low. In this paper, we propose a novel 2D Bayesian subspace analysis (2D-BSA) method for face recognition at high compression ratios. The main difference between the proposed 2D-BSA and BSA is that the former adopts a new Image-as-Matrix representation for face images, opposed to the Image-as-Vector representation in original BSA. Based on the new representation, 2D-BSA seeks two coupled set of projective vectors corresponding to the rows and columns of the difference face images, and then use them for dimensionality reduction. Experimental results on ORL and Yale face databases show that 2D-BSA is much more appropriate than BSA in recognizing faces at high compression ratios. [ABSTRACT FROM AUTHOR]

A modified ART2A-DWNN for automatic digital modulation recognition is proposed in this paper. Daubechies wavelet "db9" is chosen instead of "morlet" wavelet as the mother wavelet in ART2A-DWNN because of its compactness and orthonormality. Simulations have been carried out with the modulated signals corrupted by Gaussian noise to evaluate the performance of the proposed method. Recognition capability, noise immunity and convenience of accommodating new patterns of the modified ART2A-DWNN are simulated and analyzed. The experimental results have indicated the advantages of the modified method. Comparing the performance of the two ART2A-DWNNs, the modified ART2A-DWNN has higher recognition capability than the one with "morlet" wavelet. [ABSTRACT FROM AUTHOR]

In modern days, it is very hard to regulate cars of traffic lights violation and speed violation as well as parking violation and management of cars in parking places because of rapid increase of cars. In this paper, we proposed an intelligent recognition system of car license plates to mitigate these problems. The processing sequence of the proposed algorithm is as follows. At first, a license plate segment is extracted from an acquired car image using morphological features and color information, and noises are eliminated from the extracted license plate segment using a line scan algorithm and a grass fire algorithm, and then individual codes are extracted from the license plate segment using 4-directional edge tracking algorithm. Finally the extracted individual codes are recognized by an enhanced FCM algorithm. The enhanced FCM algorithm is a clustering algorithm improved from conventional clustering algorithms having problems that undesirable clustering results to be acquired because of distribution of patterns in cluster spaces. In order to evaluate performance of segment extraction and code recognition of the proposed method, we used 150 car images for experiment. In the results, we could verify the proposed method is more efficient and recognition performance is improved in comparison with conventional car license plate recognition methods. [ABSTRACT FROM AUTHOR]

In this paper, a new application algorithm for image recognition based on neural network has been pro-posed. The present algorithm including recognition algorithm and algorithm for training BP neural network can recognize continually changing large gray image. This algorithm has been applied to deflection measurement of bridge health monitoring, and achieved a great success. [ABSTRACT FROM AUTHOR]

In this paper, a space partition method called "Label Constrained Graph Partition" (LCGP) is presented to solve the Sample-Interweaving-Phenomenon in the original space. We first divide the entire training set into subclasses by means of LCGP, so that the scopes of subclasses will not overlap in the original space. Then "Most Discriminant Subclass Distribution" (MDSD) criterion is proposed to decide the best partition result. At last, typical LDA algorithm is applied to obtain the feature space and the RBF neural network classifier is utilized to make the final decision. The computer simulations and comparisons are given to demonstrate the performance of our method. [ABSTRACT FROM AUTHOR]

Regression is a very important data mining problem. In this paper, we present a new unbiased linear fusion method that combines component predictors so as to solve regression problems. The fusion weighted coefficients assigned are normalized, and updated by estimating the prediction errors between the component predictors and the desired regression values. The empirical results of our regression experiments on five synthetic and four benchmark data sets show that the proposed fusion method improves prediction accuracy in terms of mean-squared error, and also provides the regression curves with better fidelity with respect to normalized correlation coefficients, compared with the popular simple average and weighted average fusion rules. [ABSTRACT FROM AUTHOR]

Almost all of the well-known clustering algorithms require input parameters while these parameters may be difficult to be determined. OPTICS (Ordering Points To Identify the Clustering Structure Cluster Structure) is a primary semi-clustering method to visualize the data structure and to determine the input parameters of a given clustering algorithm. However, OPTICS has too high complexity O(n2logn) to be applied to any large dataset of n data. In this paper, we present a new semi-clustering method by partitioning data space into a number of grids and Ordering all Grids To Identify the Clustering Structure (OGTICS). Accordingly, the new method is called OGTICS. The OGTICS has only linear complexity O(n) and thus is much faster than OPTICS. Consequently, the OGTICS can be applied to very large dataset. [ABSTRACT FROM AUTHOR]

This paper proposes an orthogonal least square algorithm based on QR decomposition (QR-OLS) for the neurons selection of the hidden layer of wavelet networks. This new algorithm divides the original neurons matrix into several parts to avoid comparing among the poor ones and uses QR decomposition to select the significant ones. It can avoid lots of meaningless calculation. This algorithm is applied to the wavelet network with the analysis of variance (ANOVA) expansion and one-step-ahead predictions, respectively, for the Mackey-Glass delay-differential equation and the annual sunspot data set. The results show that the QR-OLS algorithm can relieve the load of the heave calculation and has a good performance. [ABSTRACT FROM AUTHOR]

This paper investigates a mechanism for reliable generation of sparse code in a sparsely connected, hierarchical, learning memory. Activity reduction is accomplished with local competitions that suppress activities of unselected neurons so that costly global competition is avoided. The learning ability and the memory characteristics of the proposed winner-take-all network and an oligarchy-take-all network are demonstrated using experimental results. The proposed models have the features of a learning memory essential to the development of machine intelligence. [ABSTRACT FROM AUTHOR]

We often meet the tree decomposition task in the tree kernel computing. And tree decomposition tends to vary under different tree mapping constraint. In this paper, we first introduce the general tree decomposition function, and compare the three variants of the function corresponding to different tree mapping. Then we will give a framework to generalize the kernels based on tree-to-tree decomposition with the decomposition function. [ABSTRACT FROM AUTHOR]

The implementation of multi-valued logic with a three layers forward neural network is proposed. The hidden layer is constituted by bi-threshold neurons compared with traditional simple threshold neurons. According to the obtained results in this paper, if a perception with a simple threshold neuron is used in the output layer, then the logical map of {0,1, ⋯ ,n} to {0,1} can be gained. In addition, if a linear neuron is used in the output layer, then the logical map of {0,1, ⋯ ,n} to {0,1, ⋯ ,m} can be obtained. The arithmetic, which is used to design the three layers forward neural network, improves on the traditional digital logic of two values. An example shows that the designable procedure of the network is simple and effective. [ABSTRACT FROM AUTHOR]

A major research problem in the area of unsupervised learning is the understanding of neuronal selectivity, and its role in the formation of cortical maps. Kohonen devised a self-organizing map algorithm to investigate this problem, which achieved partial success in replicating biological observations. However, a problem in using Kohonen's approach is that it does not address the stability-plasticity dilemma, as the learning rate decreases monotonically. In this paper, we propose a solution to cortical map formation which tackles the stability-plasticity problem, where the map maintains stability while enabling plasticity in the presence of changing input statistics. We adapt the parameterless SOM (Berglund and Sitte 2006) and also modify Kohonen's original approach to allow local competition in a larger cortex, where multiple winners can exist. The learning rate and neighborhood size of the modified Kohonen's method are set automatically based on the error between the local winner's weight vector and its input. We used input images consisting of lines of random orientation to train the system in an unsupervised manner. Our model shows large scale topographic organization of orientation across the cortex, which compares favorably with cortical maps measured in visual area V1 in primates. Furthermore, we demonstrate the plasticity of this map by showing that the map reorganizes when the input statistics are chanaged. [ABSTRACT FROM AUTHOR]

Based on value-at-risk (VaR) criteria, this paper presents a new class of two-stage fuzzy programming models. Because the fuzzy optimization problems often include fuzzy variables defined through continuous possibility distribution functions, they are inherently infinite- dimensional optimization problems that can rarely be solved directly. Thus, algorithms to solve such optimization problems must rely on intelligent computing as well as approximating schemes, which result in approximating finite-dimensional optimization problems. Motivated by this fact, we suggest an approximation method to evaluate critical VaR objective functions, and discuss the convergence of the approximation approach. Furthermore, we design a hybrid algorithm (HA) based on the approximation method, neural network (NN) and genetic algorithm (GA) to solve the proposed optimization problem, and provide a numerical example to test the effectiveness of the HA. [ABSTRACT FROM AUTHOR]

This paper presents the equalization of channel distortion by using a Nonlinear Neuro-Fuzzy Network (NNFN). The NFNN is constructed on the basis of fuzzy rules that incorporate nonlinear functions. The learning algorithm of NNFN is presented. The NFNN is applied for equalization of channel distortion of time-invariant and time-varying channels. The developed equalizer recovers the transmitted signal efficiently. The performance of NNFN based equalizer is compared with the performance of other nonlinear equalizers. The effectiveness of the proposed system is evaluated using simulation results of NNFN based equalization system. [ABSTRACT FROM AUTHOR]

There exist two main solutions for the classification of high-dimensional data with small number settings. One is to classify them directly in high-dimensional space with regularization methods, and the other is to reduce data dimension first, then classify them in feature space. However, which is better on earth? In this paper, the comparative studies for regularization and dimension reduction approaches are given with two typical sets of high-dimensional data from real world: Raman spectroscopy signals and stellar spectra data. Experimental results show that in most cases, the dimension reduction methods can obtain acceptable classification results, and cost less computation time. When the training sample number is insufficient and distribution is unbalance seriously, performance of some regularization approaches is better than those dimension reduction ones, but regularization methods cost more computation time. [ABSTRACT FROM AUTHOR]

We present a robust recursive least squares algorithm for multilayer feed-forward neural network training. So far, recursive least squares (RLS) has been successfully applied to training multilayer feed-forward neural networks. However, RLS method has a tendency to become diverse due to the instability in the recursive inversion procedure. In this paper, we propose a numerically robust recursive least square type algorithm using prewhitening. The proposed algorithm improves the performance of RLS in infinite numerical precision as well as in finite numerical precision. The computer simulation results in the various precision cases show that the proposed algorithm improves the numerical robustness of RLS training. [ABSTRACT FROM AUTHOR]

The increasing reliance on Computational Intelligence applications to predict stock market positions have resulted in numerous researches in financial forecasting and trading trend identifications. Stock market price prediction applications are required to be adaptive to new incoming data as well as have fast learning capabilities due to the volatility nature of market movements. This paper analyses stock market price prediction based on a Fuzzy Cerebellar Model Articulation Controller - Bayesian Ying Yang (FCMAC-BYY) neural network. The model is motivated from the Chinese ancient Ying-Yang philosophy which states that everything in the universe can be viewed as a product of a constant conflict between opposites, Ying and Yang. A perfect status is reached if Ying and Yang achieves harmony. The analyzed experiment on a set of real stock market data (Singapore Airlines Ltd - SIA) in the Singapore Stock Exchange (SGX) and Ibex35 stock index shows the effectiveness of the FCMAC-BYY in the universal approximation and prediction. [ABSTRACT FROM AUTHOR]

Product conceptual design is an innovative activity that is to form and optimize the projects of products. Identification of the best conceptual design candidate is a crucial step as design information is not complete and design knowledge is minimal at conceptual design stage. It is necessary to select the best scheme from feasible alternatives through comparison and filter. In this paper, the evaluation system of mechanism scheme is established firstly based on the performance analysis of the mechanism system and the opinions of experts. Then, the fuzzy optimum selection model of mechanism scheme evaluation is provided. Combined with the fuzzy optimum selection model with the neural network theory, a rational pattern of determining the topologic structure of network is provided. It also provides a weight-adjusted BP model of the neural network with the fuzzy optimum selection model for mechanism scheme. Finally, an example is given to verify the effective feasibility of the proposed method. [ABSTRACT FROM AUTHOR]

In this paper, a two-level decentralized decision-making problem is formulated as fuzzy random dependent-chance bilevel programming. We define the fuzzy random Nash equilibrium in the lower level problem and the fuzzy random Stackelberg-Nash equilibrium of the overall problem. In order to find the equilibria, we propose a hybrid intelligent algorithm, in which neural network, as uncertain function approximator, plays a crucial role in saving computing time, and genetic algorithm is used for optimization. Finally, we apply the fuzzy random dependent-chance bilevel programming to hierarchical resource allocation problem for illustrating the modelling idea and the effectiveness of the hybrid intelligent algorithm. [ABSTRACT FROM AUTHOR]

Many adaptive schemes for controlling the probabilities of crossover and mutation in genetic algorithms with fuzzy logic have been reported in recent years. However, there has not been known work on comparative studies of these algorithms. In this paper, several fuzzy genetic algorithms are briefly summarized first, and they are studied in comparison with each other under the same simulation conditions. The simulation results are analyzed in terms of search speed and search quality. [ABSTRACT FROM AUTHOR]

In this paper, a tabular reinforcement learning (RL) method is proposed based on improved fuzzy min-max (FMM) neural network. The method is named FMM-RL. The FMM neural network is used to segment the state space of the RL problem. The aim is to solve the "curse of dimensionality" problem of RL. Furthermore, the speed of convergence is improved evidently. Regions of state space serve as the hyperboxes of FMM. The minimal and maximal points of the hyperbox are used to define the state space partition boundaries. During the training of FMM neural network, the state space is partitioned via operations on hyperbox. Therefore, a favorable generalization performance of state space can be obtained. Finally, the method of this paper is applied to learn behaviors for the reactive robot. The experiment shows that the algorithm can effectively solve the problem of navigation in a complicated unknown environment. [ABSTRACT FROM AUTHOR]

Least squares support vector machine (LS-SVM) is a popular tool for the analysis of time series data sets. Choosing optimal hyperparameter values for LS-SVM is an important step in time series analysis. In this paper, we combine LS-SVM with simulated annealing (SA) algorithms for nonlinear time series analysis. The LS-SVM is used to predict chaotic time series, and its parameters are automatically tuned using the SA and generalization performance is estimated by minimizing the k-fold cross-validation error. A benchmark problem, Mackey-Glass time series, has been used as example for demonstration. It is showed this approach can escape from the blindness of man-made choice of the LS-SVM parameters. It enhances the prediction capability of chaotic time series. [ABSTRACT FROM AUTHOR]

In the proposed new scheme of adaptive radar target signal processing, the chaotic neural network not only detects the target signal by reconstructing the chaotic clutter, but also repairs the frequency spectrum according to its associative memory characteristic. The clutter is filtered by the Burg algorithm based on the adaptive MTI filter. The information of distance and velocity is also obtained by Burg spectral estimation. The validity of the scheme is analyzed theoretically, and the simulation results show that it has good performance in clutter and noise background. The adaptive method adopted in this paper facilitates the radar design in complex environment. [ABSTRACT FROM AUTHOR]