Rajat K. De

Feature analysis: Neural network and fuzzy set theoretic approaches

In this paper a new scheme of feature ranking and hence feature selection using a Multilayer Perception (MLP) Network has been proposed. The novelty of the proposed MLP-based scheme and its difference from another MLP-based feature ranking scheme have been analyzed. In addition we have modified an existing feature ranking/selection scheme based on fuzzy entropy. Empirical investigations show that the proposed MLP-based scheme is superior to the other schemes implemented.

Unsupervised feature selection using a neuro-fuzzy approach

A neuro-fuzzy methodology is described which involves connectionist minimization of a fuzzy feature evaluation index with unsupervised training. The concept of a flexible membership function incorporating weighed distance is introduced in the evaluation index to make the modeling of clusters more appropriate. A set of optimal weighing coefficients in terms of networks parameters representing individual feature importance is obtained through connectionist minimization. Besides, the investigation includes the development of another algorithm for ranking of different feature subsets using the aforesaid fuzzy evaluation index without neural networks. Results demonstrating the effectiveness of the algorithms for various real life data are provided.

Knowledge-based fuzzy MLP for classification and rule generation

A new scheme of knowledge-based classification and rule generation using a fuzzy multilayer perceptron (MLP) is proposed. Knowledge collected from a data set is initially encoded among the connection weights in terms of class a priori probabilities. This encoding also includes incorporation of hidden nodes corresponding to both the pattern classes and their complementary regions. The network architecture, in terms of both links and nodes, is then refined during training. Node growing and link pruning are also resorted to. Rules are generated from the trained network using the input, output, and connection weights in order to justify any decision(s) reached. Negative rules corresponding to a pattern not belonging to a class can also be obtained. These are useful for inferencing in ambiguous cases. Results on real life and synthetic data demonstrate that the speed of learning and classification performance of the proposed scheme are better than that obtained with the fuzzy and conventional versions of the MLP (involving no initial knowledge encoding). Both convex and concave decision regions are considered in the process.

Divisive Correlation Clustering Algorithm (DCCA) for grouping of genes

MOTIVATION Cluster analysis (of gene-expression data) is a useful tool for identifying biologically relevant groups of genes that show similar expression patterns under multiple experimental conditions. Various methods have been proposed for clustering gene-expression data. However most of these algorithms have several shortcomings for gene-expression data clustering. In the present article, we focus on several shortcomings of conventional clustering algorithms and propose a new one that is able to produce better clustering solution than that produced by some others. RESULTS We present the Divisive Correlation Clustering Algorithm (DCCA) that is suitable for finding a group of genes having similar pattern of variation in their expression values. To detect clusters with high correlation and biological significance, we use the correlation clustering concept introduced by Bansal et al. Our proposed algorithm DCCA produces a clustering solution without taking number of clusters to be created as an input. DCCA uses the correlation matrix in such a way that all genes in a cluster have highest average correlation with genes in that cluster. To test the performance of the DCCA, we have applied DCCA and some well-known conventional methods to an artificial dataset, and nine gene-expression datasets, and compared the performance of the algorithms. The clustering results of the DCCA are found to be more significantly relevant to the biological annotations than those of the other methods. All these facts show the superiority of the DCCA over some others for the clustering of gene-expression data. AVAILABILITY The software has been developed using C and Visual Basic languages, and can be executed on the Microsoft Windows platforms. The software may be downloaded as a zip file from http://www.isical.ac.in/~rajat. Then it needs to be installed. Two word files (included in the zip file) need to be consulted before installation and execution of the software.

Immunoinformatics: an integrated scenario

Genome sequencing of humans and other organisms has led to the accumulation of huge amounts of data, which include immunologically relevant data. A large volume of clinical data has been deposited in several immunological databases and as a result immunoinformatics has emerged as an important field which acts as an intersection between experimental immunology and computational approaches. It not only helps in dealing with the huge amount of data but also plays a role in defining new hypotheses related to immune responses. This article reviews classical immunology, different databases and prediction tools. It also describes applications of immunoinformatics in designing in silico vaccination and immune system modelling. All these efforts save time and reduce cost.

Segmentation of remotely sensed images using wavelet features and their evaluation in soft computing framework

The present paper describes a feature extraction method based on M-band wavelet packet frames for segmenting remotely sensed images. These wavelet features are then evaluated and selected using an efficient neurofuzzy algorithm. Both the feature extraction and neurofuzzy feature evaluation methods are unsupervised, and they do not require the knowledge of the number and distribution of classes corresponding to various land covers in remotely sensed images. The effectiveness of the methodology is demonstrated on two four-band Indian Remote Sensing 1A satellite (IRS-1A) images containing five to six overlapping classes and a three-band SPOT image containing seven overlapping classes.

Extraction of features using M-band wavelet packet frame and their neuro-fuzzy evaluation for multitexture segmentation

In this paper, we propose a scheme for segmentation of multitexture images. The methodology involves extraction of texture features using an overcomplete wavelet decomposition scheme called discrete M-band wavelet packet frame (DMbWPF). This is followed by the selection of important features using a neuro-fuzzy algorithm under unsupervised learning. A computationally efficient search procedure is developed for finding the optimal basis based on some maximum criterion of textural measures derived from the statistical parameters for each of the subbands. The superior discriminating capability of the extracted features for segmentation of various texture images over those obtained by several existing methods is established.

Comparing methods for metabolic network analysis and an application to metabolic engineering.

Bioinformatics tools have facilitated the reconstruction and analysis of cellular metabolism of various organisms based on information encoded in their genomes. Characterization of cellular metabolism is useful to understand the phenotypic capabilities of these organisms. It has been done quantitatively through the analysis of pathway operations. There are several in silico approaches for analyzing metabolic networks, including structural and stoichiometric analysis, metabolic flux analysis, metabolic control analysis, and several kinetic modeling based analyses. They can serve as a virtual laboratory to give insights into basic principles of cellular functions. This article summarizes the progress and advances in software and algorithm development for metabolic network analysis, along with their applications relevant to cellular physiology, and metabolic engineering with an emphasis on microbial strain optimization. Moreover, it provides a detailed comparative analysis of existing approaches under different categories.

Bi-correlation clustering algorithm for determining a set of co-regulated genes

MOTIVATION Biclustering has been emerged as a powerful tool for identification of a group of co-expressed genes under a subset of experimental conditions (measurements) present in a gene expression dataset. Several biclustering algorithms have been proposed till date. In this article, we address some of the important shortcomings of these existing biclustering algorithms and propose a new correlation-based biclustering algorithm called bi-correlation clustering algorithm (BCCA). RESULTS BCCA has been able to produce a diverse set of biclusters of co-regulated genes over a subset of samples where all the genes in a bicluster have a similar change of expression pattern over the subset of samples. Moreover, the genes in a bicluster have common transcription factor binding sites in the corresponding promoter sequences. The presence of common transcription factors binding sites, in the corresponding promoter sequences, is an evidence that a group of genes in a bicluster are co-regulated. Biclusters determined by BCCA also show highly enriched functional categories. Using different gene expression datasets, we demonstrate strength and superiority of BCCA over some existing biclustering algorithms. AVAILABILITY The software for BCCA has been developed using C and Visual Basic languages, and can be executed on the Microsoft Windows platforms. The software may be downloaded as a zip file from http://www.isical.ac.in/ approximately rajat. Then it needs to be installed. Two word files (included in the zip file) need to be consulted before installation and execution of the software. CONTACT rajat@isical.ac.in SUPPLEMENTARY INFORMATION Supplementary data are available at Bioinformatics online.

Neuro-fuzzy feature evaluation with theoretical analysis

The article provides a fuzzy set theoretic feature evaluation index and a connectionist model for its evaluation along with their theoretical analysis. A concept of weighted membership function is introduced which makes the modeling of the class structures more appropriate. A neuro-fuzzy algorithm is developed for determining the optimum weighting coefficients representing the feature importance. It is shown theoretically that the evaluation index has a fixed upper bound and a varying lower bound, and it monotonically increases with the lower bound. A relation between the evaluation index, interclass distance and weighting coefficients is established. Effectiveness of the algorithms for evaluating features both individually and in a group (considering their independence and dependency) is demonstrated along with comparisons on speech, Iris, medical and mango-leaf data. The results are also validated using scatter diagram and k-NN classifier.