Yan-Chun Liang

Solving multi objective optimization problems using particle swarm optimization

An algorithm for solving multiobjective optimization problems is presented based on PSO through the improvement of the selection manner for global and individual extremum. The search for the Pareto optimal set of multiobjective optimization problems is performed. Numerical simulations show the effectiveness of the proposed algorithm.

Data Preprocessing in SVM-Based Keywords Extraction from Scientific Documents

Scientific documents are unstructured data consisting of natural language and hard for scientists to read and manage. Keywords are very helpful for scientists to search the related documents and know about their contents in a prompt way. In this paper we investigate a kind of data preprocessing technique used in SVM-based keyword extraction from scientific documents. Four definitions of regular scientific documents are proposed, and the analysis on the experimental results is performed based on the proposed definitions. The experimental results confirm the intuition that abstract is important for keywords extraction.

Adaptive and iterative least squares support vector regression based on quadratic Renyi entropy

An adaptive and iterative LSSVR algorithm based on quadratic Renyi entropy is presented in this paper. LS-SVM loses the sparseness of support vector which is one of the important advantages of conventional SVM. The proposed algorithm overcomes this drawback. The quadratic Renyi entropy is the evaluating criterion for working set selection, and the size of working set is determined at the process of iteration adaptively. The regression parameters are calculated by incremental learning and the calculation of inversing a large scale matrix is avoided. So the running speed is improved. This algorithm reserves well the sparseness of support vector and improves the learning speed.

Diagnosis of Breast Cancer Tumor Based on ICA and LS-SVM

An efficient method for the diagnosis of breast cancer tumor is proposed based on independent component analysis (ICA) and least square support vector machine (LS-SVM). In order to save the expense of detection, firstly, variables are selected based on the theory of statistics. Then the ICA is introduced in a concise way and followed by extracting the ICA component from these selected variables. Finally the processed data are classified by the LS-SVM. Experimental and analytical results show that in the diagnosis of breast cancer tumor the proposed method is superior to the classical BP algorithm

A study on SVM with feature selection for fault diagnosis of power systems

When faults occur in power systems, it is hard to manually deal with the fault data reported by the system of supervisory control and data acquisition (SCADA) because of the huge amount of alarm information. In this paper, we study the problem of power system fault diagnosis by using support vector machine (SVM), and enhance the ability of fault diagnosis through optimizing support vectors. The results of simulation tests demonstrate the effectiveness of the proposed automatic fault diagnosis method.

Face recognition using adaptive resonance theory

Human face detection and recognition are challenged questions in pattern recognition field. After the facial features such as eyes, nose and mouth are detected in an image which contains a face, the rectangle area surrounding facial features is obtained. The pixels number of the rectangle area is large and the intensity values of these pixels are often treated as a feature vector. It is very important to drop the dimension of the vector for an effective recognition. Three means for dimensional reduction in the feature extraction field are often used, including average values of weighted intensity, wavelet transform and principle component analysis. The compact face feature vector is the eigenvector to be recognized. A face recognition method using ART2 is proposed in the paper. Experiment results show that it is preferable in recognition as well as it could increase or decrease samples rapidly.

Center selection for RBF neural network in prediction of nonlinear time series

This paper presents a new method for center selection of radial basis function (RBF) neural network. The proposed method endows a parallel quality on the process of center selection and takes advantage of the time sequential relation among time series data. Stock price prediction simulation shows that, compared with hard c-means (HCM) and orthogonal least square (OLS) RBF neural network, our method has not only better training and testing precisions, but also better generalization ability.

Electric Load Forecasting using SVMS

Support vector machines (SVMs) have been proposed as a novel technique and applied to regression recently. In this paper, SVMS are used for load forecasting. The training sample sets are chosen and preprocessed before every forecasting. Then the interference of the non-correlative and bad samples for the forecasting can be avoided. The effectiveness and the feasibility of forecasting of the employed method are examined through some simulations

Nondestructive quantitative analysis of compound paracetamol and diphenhydramine hydrochloride powder using RBF networks

A method for simultaneous analysis of the two components of compound paracetamol and diphenhydramine hydrochloride powdered drugs on near-infrared (NIR) spectroscopy is developed by using a radial basis function (RBF) network. Nearest neighbor-clustering algorithm is used as the learning algorithm of RBF network. Comparisons of the results obtained from the RBF models with those from BP models show that it is feasible to use the RBF network in nondestructive quantitative analysis of the components of drugs.

An improved line-based face recognition and indexing algorithm

Much research in human face recognition involves fronto-parallel face images, constrained rotations in and out of the plane, and under strict imaging conditions such as controlled illumination. In this paper, we propose an improved line-based face recognition and indexing algorithm that uses a set of selected rectilinear line segments of 2D face image views as the underlying image representation, together with a nearest-neighbor classifier as the line matching scheme. The algorithm achieves high generalization recognition rates for rotations both in and out of the plane and is robust to scaling and variant illumination.