Radhakrishnan Srikanth

A variable-length genetic algorithm for clustering and classification

Pattern clustering and classification can be viewed as a search for, and labeling of a set of inherent clusters in any given data set. This approach can be divided broadly into two types namely supervised and unsupervised clustering. Motivated by human perception and Kohonens method, we present a novel method of supervised clustering and classification using genetic algorithms. Clusters in the pattern space can be approximated by ellipses or sets of ellipses in two dimensions and ellipsoids in general, and the search for clusters can be approximated as the search for ellipsoids or sets of ellipsoids. By assigning fuzzy membership values to points in the pattern space a fuzzy ellipsoid is obtained. The process of thresholding which follows can be thought of as warping the contour of the ellipse to include and exclude certain points in pattern space and in effect producing an arbitrarily shaped cluster. Here we examine the use of genetic algorithms in generating fuzzy ellipsoids for learning the separation of the classes. Our evaluation function drives the genetic search towards the smallest ellipsoid or set of ellipsoids, which maximizes the number of correctly classified examples, and minimizes the number of misclassified examples.

Uncertainty management issues in the object-oriented data model

This paper fully develops a previous approach by George et al. (1993) to modeling uncertainty in class hierarchies. The model utilizes fuzzy logic to generalize equality to similarity which permitted impreciseness in data to be represented by uncertainty in classification. In this paper, the data model is formally defined and a nonredundancy preserving primitive operator, the merge, is described. It is proven that nonredundancy is always preserved in the model. An object algebra is proposed, and transformations that preserve query equality are discussed.

Fuzzy database systems : Challenges and opportunities of a new era

There have been significant theoretical advances in fuzzy database technology, yet commercially its successes have been negligible. This article examines the current state of this technology and suggests directions for future efforts. A framework for the analysis of fuzzy database technology is proposed and extant models are examined with reference to this framework. Fuzzy databases are studied in relation to the requirements of the database community. It is argued that new generation applications and object‐oriented databases hold the key to the future commercial acceptability of this technology.

A soft computing approach to intensional answering in databases

An approach to intensional answering in databases utilizing soft computing methodologies is described. The general form of a intensional answer is QYs are F, where Q is a fuzzy linguistic quantifier, Y is a class of objects, and F is a property of the class or a summary that applies to the class quantified by Q. Fuzzy descriptions of linguistic quantifiers and labels help to evaluate the degree to which an intensional answer describes a given set of tuples. Bounds on such descriptions can be defined in terms of a most general specification constituent description and a most specific generalization constraint description. A genetic algorithm technique is used to obtain near-optimal intensional answers that fit a given set of tuples.

Fuzzy elastic clustering

The authors introduce a new hybrid method called fuzzy elastic clustering for clustering and classification of patterns. It generates closed loops around a set of training patterns and sections off portions of the hyperspace thus enclosing clusters of patterns. Patterns not in the training sets are assigned fuzzy membership values with respect to each of the clusters generated. These values are defuzzified and patterns are classified as belonging to one of the clusters. This method is motivated by the elastic net approach to solving the traveling salesman problem (TSP) of R. Durbin and D. Willshaw (1987) and T. Kohonens (1982) topologically organized feature maps.<<ETX>>

Fuzzy clustering using genetic algorithms

The problem of pattern classification or clustering can be viewed as a search for a set of ellipsoids which enclose each of the clusters, presuming that, in general, clusters in the pattern space are ellipsoidal in shape. We consider fuzzy ellipsoids by assigning fuzzy membership values to patterns against each of the ellipsoids. These membership values can be defuzzified for assigning a class to the pattern. In this paper we examine the use of genetic algorithms in generating fuzzy ellipsoids for learning the separation of the classes. Our evaluation function drives the genetic search towards the smallest ellipsoid which maximizes the number of correctly classified examples, and minimizes the number of misclassified examples.<<ETX>>

Characterization of Base Periodicities in Protein-Coding Genes

A Fourier Transform of Equal Symbols (FTES) was applied as a spectral density analysis method to identify DNA bases that repeat at any frequency in selected protein-coding genes. The analysis especially focused on identification of bases responsible for the dominant signal at frequency f=1/3 found in all protein-coding genes. The study included homologous sequences from two gene families and multiple unrelated sequences from single organisms. No signal pattern or spectrum specifically characterized either gene family. However, the patterns of bases comprising the signal at f=1/3 suggested the presence of a genome-specific label for protein-coding genes from the same genome. Data suggest that three factors form the informational basis for the signal structure at f=1/3: (1) codon base positional bias; (2) codon preference; and (3) codon arrangement. Quantitative measure of the contribution of each base to the period-3 signal suggests a basis to distinguish protein-coding genes from different organisms. Application of the FTES analysis characterized genes from Escherichia coli as different from the genes from Pseudomonas aeruginosa. Preliminary analyses of genes from these and three other bacteria by artificial neural nets, using FTES parameters, support our suggestion that the period-3 informational structure contains labels for the genomic origins of protein-coding genes. FTES analysis alone or in combination with other informational measures may reveal pathways and processes of gene flow into and through natural systems of microbial cell populations.

Data transformation for learning in feedforward neural nets

The authors target the problem of curve fitting on data samples for the purpose of subsequent interpolation. This is what backpropagation was developed for but its dependency on initial conditions and net topology affects its robustness. Here the authors present a different method which is based on an analysis of possible properties of the internal representations developed as a result of learning. Thus they introduce some additional constraints concerning direct control on internal representations. This method incorporates properties of supervised as well as unsupervised learning in the fitting problem.