Sung Yang Bang

An efficient method to construct a radial basis function neural network classifier

Radial basis function neural network (RBFN) has the power of the universal function approximation. But how to construct an RBFN to solve a given problem is usually not straightforward. This paper describes a method to construct an RBFN classifier efficiently and effectively. The method determines the middle layer neurons by a fast clustering algorithm and computes the optimal weights between the middle and the output layers statistically. We applied the proposed method to construct an RBFN classifier for an unconstrained handwritten digit recognition. The experiment showed that the method could construct an RBFN classifier quickly and the performance of the classifier was better than the best result previously reported.

A handwritten numeral character classification using tolerant rough set

Proposes a data classification method based on the tolerant rough set that extends the existing equivalent rough set. A similarity measure between two data is described by a distance function of all constituent attributes and they are defined to be tolerant when their similarity measure exceeds a similarity threshold value. The determination of optimal similarity threshold value is very important for accurate classification. So, we determine it optimally by using the genetic algorithm (GA), where the goal of evolution is to balance two requirements such that: 1) some tolerant objects are required to be included in the same class as many as possible; and 2) some objects in the same class are required to be tolerant as much as possible. After finding the optimal similarity threshold value, a tolerant set of each object is obtained and the data set is grouped into the lower and upper approximation set depending on the coincidence of their classes. We propose a two-stage classification method such that all data are classified by using the lower approximation at the first stage and then the nonclassified data at the first stage are classified again by using the rough membership functions obtained from the upper approximation set. We apply the proposed classification method to the handwritten numeral character classification problem and compare its classification performance and learning time with those of the feedforward neural networks backpropagation algorithm.

Membership authentication in the dynamic group by face classification using SVM ensemble

This paper presents a method for authenticating an individuals membership in a dynamic group without revealing the individuals identity and without restricting the group size and/or the members of the group. We treat the membership authentication as a two-class face classification problem to distinguish a small size set (membership) from its complementary set (non-membership) in the universal set. In the authentication, the false-positive error is the most critical. Fortunately, the error can be validly removed by using the support vector machine (SVM) ensemble, where each SVM acts as an independent membership/non-membership classifier and several SVMs are combined in a plurality voting scheme that chooses the classification made by more than the half of SVMs. For a good encoding of face images, the Gabor filtering, principal component analysis and linear discriminant analysis have been applied consecutively to the input face image for achieving effective representation, efficient reduction of data dimension and strong separation of different faces, respectively. Next, the SVM ensemble is applied to authenticate an input face image whether it is included in the membership group or not. Our experiment results show that the SVM ensemble has the ability to recognize non-membership and a stable robustness to cope with the variations of either different group sizes or different group members. Also, we still get a reasonable membership recognition rate in spite of the limited number of membership training data.

Appearance-based gender classification with Gaussian processes

This paper concerns the gender classification task of discriminating between images of faces of men and women from face images. In appearance-based approaches, the initial images are preprocessed (e.g. normalized) and input into classifiers. Recently, support vector machines (SVMs) which are popular kernel classifiers have been applied to gender classification and have shown excellent performance. SVMs have difficulty in determining the hyperparameters in kernels (using cross-validation). We propose to use Gaussian process classifiers (GPCs) which are Bayesian kernel classifiers. The main advantage of GPCs over SVMs is that they determine the hyperparameters of the kernel based on Bayesian model selection criterion. The experimental results show that our methods outperformed SVMs with cross-validation in most of data sets. Moreover, the kernel hyperparameters found by GPCs using Bayesian methods can be used to improve SVM performance.

Face membership authentication using SVM classification tree generated by membership-based LLE data partition

This paper presents a new membership authentication method by face classification using a support vector machine (SVM) classification tree, in which the size of membership group and the members in the membership group can be changed dynamically. Unlike our previous SVM ensemble-based method, which performed only one face classification in the whole feature space, the proposed method employed a divide and conquer strategy that first performs a recursive data partition by membership-based locally linear embedding (LLE) data clustering, then does the SVM classification in each partitioned feature subset. Our experimental results show that the proposed SVM tree not only keeps the good properties that the SVM ensemble method has, such as a good authentication accuracy and the robustness to the change of members, but also has a considerable improvement on the stability under the change of membership group size.

Pattern classification using support vector machine ensemble

While the support vector machine (SVM) can provide a good generalization performance, the classification result of the SVM is often far from the theoretically expected level in practical implementation because they are based on approximated algorithms due to the high complexity of time and space. To improve the limited classification performance of the real SVM, we propose to use an SVM ensemble with bagging (bootstrap aggregating) or boosting. In bagging, each individual SVM is trained independently, using randomly chosen training samples via a bootstrap technique. In boosting, each individual SVM is trained using training samples chosen according to the samples probability distribution, which is updated in proportion to the degree of error of the sample. In both bagging and boosting, the trained individual SVMs are aggregated to make a collective decision in several ways, such as majority voting, least squares estimation based weighting, and double-layer hierarchical combination. Various simulation results for handwritten digit recognition and fraud detection show that the proposed SVM ensemble with bagging or boosting greatly outperforms a single SVM in terms of classification accuracy.

Support Vector Machine Ensemble with Bagging

Even the support vector machine (SVM) has been proposed to provide a good generalization performance, the classification result of the practically implemented SVM is often far from the theoretically expected level because their implementations are based on the approximated algorithms due to the high complexity of time and space. To improve the limited classification performance of the real SVM, we propose to use the SVM ensembles with bagging (bootstrap aggregating). Each individual SVM is trained independently using the randomly chosen training samples via a bootstrap technique. Then, they are aggregated into to make a collective decision in several ways such as the majority voting, the LSE(least squares estimation)-based weighting, and the double-layer hierarchical combining. Various simulation results for the IRIS data classification and the hand-written digit recognitionshow that the proposed SVM ensembles with bagging outperforms a single SVM in terms of classification accuracy greatly.

Face recognition using the mixture-of-eigenfaces method

This paper deals with face recognition using the mixture-of-eigenfaces method. The well-known eigenface method uses one set of holistic facial features obtained by principal component analysis (PCA). However, a single set of eigenfaces is not enough to represent face images with large variations. To overcome this weakness, we propose the mixture-of-eigenfaces method, which uses more than one set of eigenfaces obtained from the expection maximization learning in the PCA mixture model. In this method, several sets of eigenfaces are obtained from all face images, and each template face image is represented by an appropriate set of eigenfaces. Recognition was performed using the distance between the input image and the labelled template image stored in the face database, where the distance is the difference of the feature values that are obtained from the set of eigenfaces indicated by the labelled template image. Simulation results show that the proposed mixture-of-eigenfaces method outperforms the eigenface method in terms of recognition accuracy for face images with pose and illumination variations.

Handwritten Korean character image database PE92

The purpose of the current PE92 database project is two fold. One is to provide a comprehensive set of character image data to a developer of a recognition system so that the developer can concentrate on developing an algorithm. The other is to offer a means by which an evaluator can compare various algorithms objectively. The authors collected 100 sets of KS 2350 handwritten Korean character images. They tried to collect as many writing styles as possible. The first 70 sets were generated by more than 500 different writers, and each of the remaining 30 sets was written by the same person. Writers wrote down the characters in prespecified boxes and the database was created by scanning the data sheets by an image scanner. Each image is the size of 100/spl times/100 with 256 gray levels. Finally, the authors analyze the quality of the database created and calculated various statistics of the database PE92.<<ETX>>

Face recognition using LDA mixture model

Linear discriminant analysis (LDA) provides the projection that discriminates data well, and shows a good performance for face recognition. However, since LDA provides only one transformation matrix over the whole data, it is not sufficient to discriminate complex data consisting of many classes with high variations, such as human faces. To overcome this weakness, we propose a new face recognition method based on the LDA mixture model, where the set of all classes are partitioned into several clusters and we obtain a transformation matrix for each cluster. This accurate and detailed representation will improve classification performance. Simulation results of face recognition show that LDA mixture model outperforms PCA, LDA, and PCA mixture model in terms of classification performance.