International Journal of Pattern Recognition and Artificial Intelligence

We propose a new method to categorize continuous numeric percepts for Q-learning, where percept vectors are classified into categories on the basis of fuzzy ART and Q-learning uses categories as states to acquire rules for agent behavior. For efficient learning, we modify fuzzy ART to reduce the number of categories without deteriorating the efficiency of reinforcement learning. In our modification, a vigilance parameter is defined for each category in order to control the size of a category and it is updated during learning. The method to update a vigilance parameter is based on category integration, which contributes to reducing the number of categories. Here, we define the similarity for any category pair to judge whether category integration should be performed or not. When two categories are integrated into a new category, a vigilance parameter for the category is calculated and categories used for integration are discarded, so that the number of categories is reduced without restricting the number of categories. Experimental results show that Q-learning with modified fuzzy ART acquires good rules for agent behavior more efficiently than Q-learning with ordinary fuzzy ART, although the number of categories generated by modified fuzzy ART is much less than that generated by ordinary one.

In this paper, a modified learning algorithm to obtain better generalization performance is proposed. The cost terms of this new algorithm are selected based on the second-order derivatives of the neural activation at the hidden layers and the first-order derivatives of the neural activation at the output layer. It can be guaranteed that in the course of training, the additional cost terms for this algorithm can penalize both the input-to-output mapping sensitivity and the high frequency components to obtain better generalization performance. Finally, theoretical justifications and simulation results are given to verify the efficiency and effectiveness of the proposed learning algorithm.

A novel methodology is proposed for the development of neural network models for complex engineering systems exhibiting nonlinearity. This method performs neural network modeling by first establishing some fundamental nonlinear functions from a priori engineering knowledge, which are then constructed and coded into appropriate chromosome representations. Given a suitable fitness function, using evolutionary approaches such as genetic algorithms, a population of chromosomes evolves for a certain number of generations to finally produce a neural network model best fitting the system data. The objective is to improve the transparency of the neural networks, i.e. to produce physically meaningful "white box" neural network model with better generalization performance. In this paper, the problem formulation, the neural network configuration, and the associated optimization software are discussed in detail. This methodology is then applied to a practical real-world system to illustrate its effectiveness.

The study on neural systems is very hot, especially regarding modelling of fuzzy neural networks. The neuron models have been limited in interval [0,1] before. This paper studies the logic operators and neuron models of proposition object based on [a, b]. Any interval [a, b] is called a generalized interval. Firstly, authors provide the conception of proposition object, discussed on the radix space of universal logic. Secondly, using the NTS norms theories, this paper establishes the universal logic operation models on generalized interval [a, b]. Thirdly, the paper builds up a new uniform neuron model of "Not/And/Or/Average" on generalized interval [a, b]. As an instance, authors discuss the neuron models based on standard interval [0,1], which are continuously changeable with generalized correlation coefficient "h" and generalized self-correlation coefficient "k". This work offers important theories and models for neurons reasoning, make it more flexible because of the change of h, k and [a, b], and enlarge its study domain.

Term structure is a useful curve describing some financial asset as a function of time to maturity or expiration. In this paper, we propose to use Independent Component Analysis (ICA) to model the term structure of multiple yield curves. The idea is that we first employ ICA to decompose the multivariate time series, then we suggest two ICA methods for dimension reduction and pattern recognition of the term structure. We also compare the results by using an alternative method, Principal Component Analysis (PCA). The empirical studies suggest that the proposed ICA approaches outperform PCA methods in modeling the term structure. This model can be used in financial time series analysis as well as related financial applications.

In face recognition tasks, Fisher discriminant analysis (FDA) is one of the promising methods for dimensionality reduction and discriminant feature extraction. The objective of FDA is to find an optimal projection matrix, which maximizes the between-class-distance and simultaneously minimizes within-class-distance. The main limitation of traditional FDA is the so-called Small Sample Size (3S) problem. It induces that the within-class scatter matrix is singular and then the traditional FDA fails to perform directly for pattern classification. To overcome 3S problem, this paper proposes a novel two-step single parameter regularization Fisher discriminant (2SRFD) algorithm for face recognition. The first semi-regularized step is based on a rank lifting theorem. This step adjusts both the projection directions and their corresponding weights. Our previous three-to-one parameter regularized technique is exploited in the second stage, which just changes the weights of projection directions. It is shown that the final regularized within-class scatter matrix approaches the original within-class scatter matrix as the single parameter tends to zero. Also, our method has good computational complexity. The proposed method has been tested and evaluated with three public available databases, namely ORL, CMU PIE and FERET face databases. Comparing with existing state-of-the-art FDA-based methods in solving the S3 problem, the proposed 2SRFD approach gives the best performance.

Handwriting-based personal identification, which is also called handwriting-based writer identification, is an active research topic in pattern recognition. Despite continuous effort, offline handwriting-based writer identification still remains as a challenging problem because writing features can only be extracted from the handwriting image. As a result, plenty of dynamic writing information, which is very valuable for writer identification, is unavailable for offline writer identification. In this paper, we present a novel wavelet-based Generalized Gaussian Density (GGD) method for offline writer identification. Compared with the 2-D Gabor model, which is currently widely acknowledged as a good method for offline handwriting identification, GGD method not only achieves a better identification accuracy but also greatly reduces the elapsed time on calculation in our experiments.

Recently, a new personalized model has been developed to prevent credit card fraud. This model is promising; however, there remains some problems. Existing approaches cannot identify well credit card frauds from few data with skewed distributions. This paper proposes to address the problem using a binary support vector system (BSVS). The proposed BSVS is based on the support vectors in the support vector machines (SVM) and the genetic algorithm (GA) is employed to select support vectors. To obtain a high true negative rate, self-organizing mapping (SOM) is first employed to estimate the distribution model of the input data. Then BSVS is used to best train the data according to the input data distribution to obtain a high detection rate. Experimental results show that the proposed BSVS is effective especially for predicting a high true negative rate.

This paper proposes document clustering using Kohonen neural network with pairwise constraints. This algorithm works directly on textual information without mapping document into some representation that has quantitative features. The input level of the proposed neural network can directly receive a qualitative value without mapping the qualitative value into the numerical value. The proposed neural network is based on the architecture of text processing Kohonen neural network, the concepts of dissimilarity measure of symbolic objects and pairwise constrained concepts. As a result, the model can successfully assign cluster label to the objects.

In order to improve the performance of mass segmentation on mammograms, an intelligent algorithm is proposed in this paper. It establishes two mass models to characterize the various masses, and the ones in the denser tissue are represented with Model I, while the ones in the fatty tissue are represented with Model II. Then, it uses iterative thresholding to extract the suspicious area, as well as the rough regions of those masses matching Model II, and applies a DWT-based technique to locate those masses matching Model I, which are hidden in the high gray-level intensity and contrast area. A region growing process restricted by Canny edge detection is subsequently used to segment the rough regions of those masses matching Model I, and finally snakes are carried out to find all the mass regions roughly extracted above. Thirty patient cases with 60 mammograms and 107 masses were used for evaluation, and the experimental result has demonstrated the algorithm's better performance over the conventional methods.

A new method for selecting features from protein sequences is proposed in this paper. First, the protein sequences are converted into fixed-dimensional feature vectors. Then, a subset of features is selected using relative entropy method and used as the inputs for Support Vector Machine (SVM). Finally, the trained SVM classifier is utilized to classify protein sequences into certain known protein families. Experimental results over proteins obtained from PIR database and GPCRs have shown that our proposed approach is really effective and efficient in selecting features from protein sequences.

In the post-genomic era, predicting protein function is a challenging problem. It is difficult and burdensome work to unravel the functions of a protein by wet experiments only. In this paper, we propose a novel method to predict protein functions by building a "Protein Interaction Network Dictionary (PIND)". This method deduces the protein functions by searching the most similar "words"(an anagram of functions in neighbor proteins on a protein–protein interaction graph) using global alignments. An evaluation of sensitivity and specificity shows that this PIND approach outperforms previous approaches such as Majority Rule and Chi-Square measure, and that it competes with the recently introduced Random Markov Model approach.

This paper proposes a new method for tumor classification using gene expression data, which mainly contains three steps. Firstly, the original DNA microarray gene expression data are selected using t-statistics. Secondly, the selected genes are modeled by Independent Component Analysis (ICA). Finally, Support Vector Machine (SVM) is used to classify the modeling data. To show the validity of the proposed method, we apply it to classify two DNA microarray data sets involving various human normal and tumor tissue samples. The experimental results show that the method is efficient and feasible.

There is great potential in mining gene expression microarray databases to discover causal relationships in the gene-regulation pathway. Several methods using Bayesian networks have been reported. Most of them use a heuristics search based on the criteria for choosing a network, but these methods are often computationally intractable for microarray data with thousands of genes. In this work, a simple constrained-based, local causal discovery method is presented. This method is computationally feasible but does not attempt to discover complete causal structure.

To show the effectiveness of this method, we have conducted simulations and applied this method to the data set from Hughes et al. from 300 expression profiles of yeast. Using this method, results of simulation data tests demonstrated that the accuracy ratios of causal relationships became higher when the sample size increased. From the yeast data set, a number of causal relations were found. A cursory analysis shows some of the relations have biological sense, others need further investigation.