In this paper, we present a self-organization neural network approach for spatial data visualization and spatial data indexing. Spatial data is typically used to represent multi-dimensional objects. Generally, for efficient processing such as indexing and retrieval, each multi-dimensional object is represented by an isothetic minimum bounding rectangle. Direct visualization of these multi-dimensional rectangles, denoting spatial objects, is not possible, if the number of dimensions exceeds three. Many linear and non-linear mapping techniques have been proposed in the literature for mapping point data, i.e., data that are points in multi-dimensional space. These approaches map points in higher-dimensional space to lower-dimensional space. Making use of these point data mapping approaches is a computationally intensive task as the number of points to be mapped is very large. In this paper, we propose a Kohonen's self-organization neural network approach for clustering spatial data. Cluster prototypes associated with nodes in the network are mapped into lower dimensions for data visualization using a non-linear mapping technique. We explain the applicability of this approach for efficient indexing of spatial data.

This paper presents a clan-based evolutionary approach for solving control problems. Three selected control problems, viz. linear-quadratic, harvest, and push-cart problems, are solved using the proposed approach. Results are compared with those of the evolutionary programming (EP) approach. In most of the cases, the proposed approach is successful in obtaining (near) optimal solutions for these selected problems.

In this paper, we describe the applicability of the K-means clustering algorithm for locating thresholds in a given histogram. In order to find optimal thresholds a probabilistic method called Multi-state Stochastic Connectionist Approach (MSCA) is employed. Mean Field Annealing (MFA), a deterministic counterpart of MSCA, is also studied in this context. A parallel model to parallelize the above methods is presented. Results of MFA and MSCA are compared with that of the K-means algorithm.

The applicability of evolution strategies (ESs), population based stochastic optimization techniques, to optimize clustering objective functions is explored. Clustering objective functions are categorized into centroid and non-centroid type of functions. Optimization of the centroid type of objective functions is accomplished by formulating them as functions of real-valued parameters using ESs. Both hard and fuzzy clustering objective functions are considered in this study. Applicability of ESs to discrete optimization problems is extended to optimize the non-centroid type of objective functions. As ESs are amenable to parallelization, a parallel model (master/slave model) is described in the context of the clustering problem. Results obtained for selected data sets substantiate the utility of ESs in clustering.

This paper presents a stochastic connectionist approach for cluster analysis. Clustering problem is formulated as a real-parameter function optimization problem and is solved using the proposed approach. As the proposed connectionist approach performs stochastic search, it avoids getting stuck in a local minimum, and guarantees asymptotic convergence to optimal solution. The amenability of connectionist approaches to massive parallelization enables one to obtain linear speedup with available parallel hardware. Several data sets are clustered using the proposed approach and the partitions obtained are (near) optimal in nature. Results pertaining to some important data sets are presented.< >

In this paper, we explore the applicability of simulated annealing, a probabilistic search method, for finding optimal partition of the data. A new formulation of the clustering problem is investigated. In order to obtain optimal partition, search is undertaken to locate optimal initial seeds, such that the K-means algorithm converges to optimal partition. Search space involved in this process is continuous, so decretization is done and simulated annealing is employed for locating optimal initial seeds. Experimental results substantiate the proposed method. Results obtained with the selected data sets are presented.

A connectionist approach for global optimization is proposed. The standard function set is tested. Results obtained, in the case of large scale problems, indicate excellent scalability of the proposed approach

Designing good error-correcting codes typically requires searching in search spaces. The vastness of search space precludes the use of brute force techniques such as exhaustive enumeration. The problem of designing codes so that each code repels others (in the sense of hamming distance) fits well in the framework of neural networks. Formulating an energy function to design codes is very difficult and cannot satisfactorily be solved by Hopfield neural network model. To alleviate these problems, a probabilistic neural network model is proposed. The usefulness of the proposed model is investigated with respect to maximal distance codes and constant weight codes. Results of some code parameters that have been designed using the proposed model are presented.