T. Kathirvalavakumar

Reverse Engineering the Neural Networks for Rule Extraction in Classification Problems

Artificial neural networks often achieve high classification accuracy rates, but they are considered as black boxes due to their lack of explanation capability. This paper proposes the new rule extraction algorithm RxREN to overcome this drawback. In pedagogical approach the proposed algorithm extracts the rules from trained neural networks for datasets with mixed mode attributes. The algorithm relies on reverse engineering technique to prune the insignificant input neurons and to discover the technological principles of each significant input neuron of neural network in classification. The novelty of this algorithm lies in the simplicity of the extracted rules and conditions in rule are involving both discrete and continuous mode of attributes. Experimentation using six different real datasets namely iris, wbc, hepatitis, pid, ionosphere and creditg show that the proposed algorithm is quite efficient in extracting smallest set of rules with high classification accuracy than those generated by other neural network rule extraction methods.

Rule extraction from neural networks — A comparative study

Though neural networks have achieved highest classification accuracy for many classification problems, the obtained results may not be interpretable as they are often considered as black box. To overcome this drawback researchers have developed many rule extraction algorithms. This paper has discussed on various rule extraction algorithms based on three different rule extraction approaches namely decompositional, pedagogical and eclectic. Also it evaluates the performance of those approaches by comparing different algorithms with these three approaches on three real datasets namely Wisconsin breast cancer, Pima Indian diabetes and Iris plants.

A Novel Pruning Algorithm for Optimizing Feedforward Neural Network of Classification Problems

Optimizing the structure of neural networks is an essential step for the discovery of knowledge from data. This paper deals with a new approach which determines the insignificant input and hidden neurons to detect the optimum structure of a feedforward neural network. The proposed pruning algorithm, called as neural network pruning by significance (N2PS), is based on a new significant measure which is calculated by the Sigmoidal activation value of the node and all the weights of its outgoing connections. It considers all the nodes with significance value below the threshold as insignificant and eliminates them. The advantages of this approach are illustrated by implementing it on six different real datasets namely iris, breast-cancer, hepatitis, diabetes, ionosphere and wave. The results show that the proposed algorithm is quite efficient in pruning the significant number of neurons on the neural network models without sacrificing the networks performance.

Pruning algorithms of neural networks — a comparative study

The neural network with optimal architecture speeds up the learning process and generalizes the problem well for further knowledge extraction. As a result researchers have developed various techniques for pruning the neural networks. This paper provides a survey of existing pruning techniques that optimize the architecture of neural networks and discusses their advantages and limitations. Also the paper evaluates the effectiveness of various pruning techniques by comparing the performance of some traditional and recent pruning algorithms based on sensitivity analysis, mutual information and significance on four real datasets namely Iris, Wisconsin breast cancer, Hepatitis Domain and Pima Indian Diabetes.

A new discretization algorithm based on range coefficient of dispersion and skewness for neural networks classifier

In this paper we propose a new static, global, supervised, incremental and bottom-up discretization algorithm based on coefficient of dispersion and skewness of data range. It automates the discretization process by introducing the number of intervals and stopping criterion. The results obtained using this discretization algorithm show that the discretization scheme generated by the algorithm almost has minimum number of intervals and requires smallest discretization time. The feedforward neural network with conjugate gradient training algorithm is used to compute the accuracy of classification from the data discretized by this algorithm. The efficiency of the proposed algorithm is shown in terms of better discretization scheme and better accuracy of classification by implementing it on six different real data sets.

Efficient Touching Text Line Segmentation in Tamil Script Using Horizontal Projection

In this paper an efficient method has been proposed to segment a document of machine printed Tamil sources into text lines. Because of the interfering lines, text line segmentation remain a problem. Standard Horizontal projection method can not segment the lines which are overlapped or touched. But the proposed method uses horizontal projection technique to solve the problem of line overlapping and over segmentation. Experimental results show that 100% accuracy is obtained from the line segmentation process which involves Tamil language document with different sizes and different fonts with line overlapping.

Features Reduction using Wavelet and Discriminative Common Vector and Recognizing Faces using RBF

Recognizing patterns by radial basis function network using reduced features obtained through wavelet transformation and discriminative common vector is proposed. Wavelet coefficients obtained after applying wavelet transformations on input patterns, are used to extract significant features from the samples. The discriminative common vectors are extracted using the within-class scatter matrix method from the wavelet coefficients. The discriminative common vectors are classified using radial basis function network. The proposed system is validated using three face databases such as ORL, The Japanese Female Facial Expression (JAFFE) and Essex Face database. The proposed method reduces the number of features, minimizes the computational complexity and provides better recognition rates.

Face Recognition by RBF with Wavelet, DCV and Modified LBP Operator Face Representation Methods

A face recognition system must be robust with respect to many variability such as viewpoint, illumination, and facial expression of the face image. The main aim of the proposed work is to represent and recognize face images with different poses. An efficient face recognition system with face image representation using wavelet and averaged wavelet packet coefficients in the form of Discriminative Common Vector (DCV) and modified Local Binary Patterns (LBP) and recognition using radial basis function (RBF) neural network is presented. Face images are decomposed by 2-level two-dimensional (2-D) wavelet and wavelet packet transformation. The discriminative common vectors are obtained for wavelet and averaged wavelet packet coefficients. Newly proposed LBP operator is applied on the DCV and LBPs are obtained. Histogram values are generated for the LBP and recognized using RBF network. The proposed work is tested on three standard face databases namely Olivetti-Oracle Research Lab (ORL), Japanese Female Facial Expression (JAFFE) and Essex face database. The extracted features are recognized by the proposed method results in good recognition rates. The execution time for the proposed methods is also less because of the meaningful extracted features obtained from the face representation methods.

Efficient Handwritten Numeral Recognition System Using Leaders of Separated Digit and RBF Network

In this paper an efficient method has been proposed to classify handwritten numerals using leader algorithm and Radial Basis Function network. Handwritten numerals are represented in matrix form and clusters with leaders are formed for each row of each digit separately. Every leader is with single target digit. Duplication patterns are avoided from the cluster leaders by combining those in a single pattern with target vectors having corresponding bits in on mode. Now resultant target vectors are with 10 bits corresponding to the number of digits considered for classification. Constructed leaders are trained using Radial Basis Function network. Experimental results show that the minimum number of patterns are enough for training compared to total patterns and it has been observed that convergency is fast during training. Also the number of resultant leaders after avoiding duplication patterns are less and the number of bits in each resultant pattern is 12.

Face Representation Using Averaged Wavelet, Micro Patterns and Recognition Using RBF Network

Recognition of human faces is a very important task in many applications such as authentication and surveillance. An efficient face recognition system with face image representation using averaged wavelet and wavelet packet coefficients, Discriminative Common Vector (DCV) and modified Local Binary Patterns (LBP) and recognition using radial basis function (RBF) network is presented. Face images are decomposed by 2-level wavelet and wavelet packet transformation. The discriminative common vectors are obtained for averaged wavelet. The new proposed LBP operator is applied on the obtained DCV and also applied on averaged wavelet packet coefficients of all the samples of a class. The histogram values obtained from the LBP are recognized using RBF network. The proposed work is tested on three face databases such as Olivetti Oracle Research Lab (ORL), Japanese Female Facial Expression (JAFFE) and Essex face database. The proposed method results in good recognition rates along with less training time because of the extracted discriminant input from the preprocessing steps involved in the proposed work.