M. Arif Wani

Parallel Edge-Region-Based Segmentation Algorithm Targeted at Reconfigurable MultiRing Network

In this paper, we present the parallel edge-region-based segmentation algorithm targeted at reconfigurable MultiRing network. The algorithm is based on detection of edges in the image. The 3-D image is sliced to create equidepth contours (EDCs). Three types of critical points, corresponding to three types of edges: fold edges, semistep edges, and boundary edges, are extracted in parallel from various EDCs. A subset of edge pixels is extracted first using these critical points. The edges are grown in parallel from these pixels through application of edge masks. The parallel algorithm is targeted on the MultiRing network. Various broadcasting mechanisms for utilizing the MultiRing for various stages of the algorithm are discussed. The paper also discusses how the segmentation algorithm is mapped on the MultiRing topology.

Introducing Subspace Grids to Recognise Patterns in Multidimensinal Data

The work presented in this paper proposes a new approach of using subspace grids for recognizing patterns in multidimensional data. The proposed approach addresses the two problems often associated with this task: i) curse of dimensionality ii) cases with small sample sizes. To handle the curse of dimensionality problem, this paper introduces subspace grids and shows how it can be employed for pattern recognition tasks efficiently. To address the cases with small sample sizes, this paper proposes a multi-scale approach where coarse scale, being stable and generic in nature, suits well for small sample sizes, and fine scales, being more specialized in nature, enhance classification accuracy. The paper first describes projection of multidimensional data to a number of lower dimensional subspaces. Principal component analysis (PCA) and multiple discriminant analysis (MDA) algorithms are used to define lower dimensional subspaces. The range of value associated with each vector of a subspace is divided into a number of equal parts to define coarse subspace grids. Coarse subspace grids are further divided equally into fine subspace grids. A recursive procedure is then employed to obtain rules where coarse and fine subspace grids form premises of rules. The system is tested on the bench mark IRIS data set having 150 examples. (50 examples belonging to each class type). The results show that the use of subspaces grids produces good results to recognize patterns in multidimensional data.

Microarray Classification Using Sub-space Grids

The work presented in this paper describes how sub-space grids can be employed to extract rules for micro array classification. The paper first describes principal component analysis (PCA) algorithm for obtaining sub-space grids from the projected low dimensional space. A recursive procedure is then used to obtain rules where sub-space grids form premises of rules. The extracted set of rules is evaluated on both training and testing data sets. The sub-space grids from PCA algorithm are characterized by overlapped data from different classes and use of even more than two premises in a rule does not fully address the problem of overlapped data. As such the rules obtained do not discriminate different classes accurately. To increase the effectiveness of the set of rules, multiple discriminant analysis (MDA) algorithm instead of PCA algorithm is employed to obtain sub-space grids from the projected low dimensional space. These sub-space grids from MDA algorithm improve the classification accuracy of the system. However, the size of set of rules extracted is large and these rules are sensitive to local variations associated with the data. To address these issues, the paper explores using both the PCA and MDA algorithms simultaneously fo projected low dimensional space for obtaining sub-space grids. The resulting set of rules produce better classification accuracy results. The paper discusses a comprehensive evaluation of this rule based system. The system is tested on a dataset of 62 samples (40 colon tumor and 22 normal colon tissue). The results show that the use of sub-space grids that are obtained from a projected low dimensional space of combined PCA and MDA algorithms increase the accuracy of classification results of micro array data.

Local and Global Data Spread Based Index for Determining Number of Clusters in a Dataset

Most of the clustering algorithms are sensitive to the input parameters and produce different clustering results for different input parameters for same datasets. A number of methods and indices have been proposed for validating results of a clustering process. The most commonly used approaches for cluster validation are based on internal indices. In this paper, we propose a new cluster validity index (ARSpread index) for the purpose of cluster validation and determining number of clusters present in a dataset. Local and global data spread based approach is proposed to measure the compactness of a cluster. A distinctness measure that is based on a penalty function is incorporated in the proposed index. We conduct a thorough comparison of five commonly known indices with the proposed index and provide a summary of experimental performance of different indices. Experimental results show that the proposed new index performs better than the commonly known indices.

A new cluster validity index using maximum cluster spread based compactness measure

Purpose – The most commonly used approaches for cluster validation are based on indices but the majority of the existing cluster validity indices do not work well on data sets of different complexities. The purpose of this paper is to propose a new cluster validity index (ARSD index) that works well on all types of data sets. Design/methodology/approach – The authors introduce a new compactness measure that depicts the typical behaviour of a cluster where more points are located around the centre and lesser points towards the outer edge of the cluster. A novel penalty function is proposed for determining the distinctness measure of clusters. Random linear search-algorithm is employed to evaluate and compare the performance of the five commonly known validity indices and the proposed validity index. The values of the six indices are computed for all nc ranging from (nc min, nc max) to obtain the optimal number of clusters present in a data set. The data sets used in the experiments include shaped, Gaussian...

Hybrid Neural Network Based Model for Predicting the Performance of a Two Stroke Spark Ignition Engine

This paper describes a hybrid neural network based model for predicting the performance of a single cylinder two stroke cycle spark ignition engine. The engine was run in the carburetor mode and engine mapping was done by collecting the engine performance data in terms of power and brake specific fuel consumption for various combinations of speed, load and air-fuel ratio. This data was used for predicting the engine performance. The work first presents a model that is based on conventional thermodynamic and gas dynamic relations. The performance of the model is improved by integrating a conventional model with a distributed and synergistic neural network. The resulting hybrid model follows closely the expected results in predicting the performance of a two stroke cycle spark ignition engine. The analysis shows that the hybrid model has learnt the input output data relation very well and is capable to predict the output in the decided domain.With the rapid growth of the World Wide Web (www), it becomes a critical issue to design and organize the vast amounts of on-line documents on the web according to their topic. Even for the search engines it is very important to group similar documents in order to improve their performance when a query is submitted to the system. Clusterng is useful for taxonomy design and similarity search of documents on such a domain. Similarity is fundamental to many clustering applications on hypertext. In this paper, we will study how measures of similarity are used to cluster a collection of documents on a web site. Most of the document clustering techniques rely on single term analysis of text, such as vector space model. To better group of related documents we propose a new semantic similarity measure. We compare our measure with Wu-Palmer similarity and cosine similarity. Experimental results show that cosine similarity perform better than the semantic similarities. We demonstrate our results on Turkish documents. This is a first study that considers the semantic similarities between Turkish documents.

A novel point density based validity index for clustering gene expression datasets

Elucidating the patterns hidden in gene expression data offers an opportunity for identifying co-expressed genes and biologically relevant grouping of genes. However, the large number of genes and the complexity of biological networks greatly increase the challenges of comprehending and interpreting the microarray data. A first step toward addressing this challenge is the use of clustering techniques. Validation of results obtained from a clustering algorithm is an important part of the clustering process. In this paper, we propose a new cluster validity index (ARPoints index) for the purpose of cluster validation. A new approach to determine the compactness measure and distinctness measure of clusters is presented. We revisit commonly known indices and conduct a thorough comparison of these indices with the proposed index and provide a summary of performance evaluation of different indices. Experimental results show that the proposed index performs better than the commonly known cluster validity indices.

Efficient and Rotation Invariant Fingerprint Matching Algorithm Using Adjustment Factor

This paper presents a new efficient and rotation invariant algorithm that makes use of local features forfingerprint matching. Minutiae points are first extracted from afingerprint image. Minutiae code mc, defined in this paper, is then generated for each extracted minutiae point. The proposed minutiae code is invariant to rotation of the fingerprint image. Adjustment factor (AF) is introduced to address the problem due to differences in a claimant fingerprint and a template fingerprint of the same person that may be present due to variations in inking or variations in pressure applied between a finger and the scanner. Adjustment factor is calculated from the minutiae code (mc) of the two fingerprints being matched. A two stage fingerprint matching process is proposed. During first stage only a few minutiae codes are checked to decide if the second stage of matching process is required. This makes the matching process faster. The proposed strategy is tested on a number of publicly available images (DB1 of FVC2004 database) and the results are promising.

Performance Comparison of Major Classical Face Recognition Techniques

The goal of this paper is to present a critical comparison of existing classical techniques on recognition of human faces. This paper describes the four major classical face recognition techniques i.e., i) Principal Component Analysis (PCA), ii) Linear Discriminant Analysis (LDA), iii) Discrete Cosine Transform (DCT), and iv) Independent Component Analysis (ICA). Strong and weak features of these techniques are discussed. The paper then provides performance comparison and a generalized discussion of the training requirements for these face recognition techniques. Extensive experimental results with three publicly available databases (ORL, Yale, FERET databases) are provided. Performance comparison of recognizing face images taken under varying facial expressions, varying lighting condition and varying poses are discussed.

Classification of Microarray Data Using Subspace Grids with Synergistic and Distributed Neural Network Models

Synergistic and distributed neural network models are employed in this work for Microarray data classification. The proposed approach uses subspace grids as input to synergistic and distributed neural network models. The paper first describes projection of multidimensional Microarray data to a number of lower dimensional subspaces. This work makes use of two algorithms to define lower dimensional subspaces. The range of value associated with each vector of a subspace is divided into a number of equal parts to define subspace grids. The resulting subspace grid data is used with the proposed synergistic and distributed neural network models to classify patterns associated with multidimensional Microarray data. The results show that the use of subspaces grids with synergistic and distributed neural network models produces good results to classify patterns in multidimensional Microarray data.