Partha Garai

Fuzzy–Rough Simultaneous Attribute Selection and Feature Extraction Algorithm

Among the huge number of attributes or features present in real-life data sets, only a small fraction of them are effective to represent the data set accurately. Prior to analysis of the data set, selecting or extracting relevant and significant features is an important preprocessing step used for pattern recognition, data mining, and machine learning. In this regard, a novel dimensionality reduction method, based on fuzzy-rough sets, that simultaneously selects attributes and extracts features using the concept of feature significance is presented. The method is based on maximizing both the relevance and significance of the reduced feature set, whereby redundancy therein is removed. This paper also presents classical and neighborhood rough sets for computing the relevance and significance of the feature set and compares their performances with that of fuzzy-rough sets based on the predictive accuracy of nearest neighbor rule, support vector machine, and decision tree. An important finding is that the proposed dimensionality reduction method based on fuzzy-rough sets is shown to be more effective for generating a relevant and significant feature subset. The effectiveness of the proposed fuzzy-rough-set-based dimensionality reduction method, along with a comparison with existing attribute selection and feature extraction methods, is demonstrated on real-life data sets.

On fuzzy-rough attribute selection: Criteria of Max-Dependency, Max-Relevance, Min-Redundancy, and Max-Significance

Attribute selection is one of the important problems encountered in pattern recognition, machine learning, data mining, and bioinformatics. It refers to the problem of selecting those input attributes or features that are most effective to predict the sample categories. In this regard, rough set theory has been shown to be successful for selecting relevant and nonredundant attributes from a given data set. However, the classical rough sets are unable to handle real valued noisy features. This problem can be addressed by the fuzzy-rough sets, which are the generalization of classical rough sets. A feature selection method is presented here based on fuzzy-rough sets by maximizing both relevance and significance of the selected features. This paper also presents different feature evaluation criteria such as dependency, relevance, redundancy, and significance for attribute selection task using fuzzy-rough sets. The performance of different rough set models is compared with that of some existing feature evaluation indices based on the predictive accuracy of nearest neighbor rule, support vector machine, and decision tree. The effectiveness of the fuzzy-rough set based attribute selection method, along with a comparison with existing feature evaluation indices and different rough set models, is demonstrated on a set of benchmark and microarray gene expression data sets.

IT2 Fuzzy-Rough Sets and Max Relevance-Max Significance Criterion for Attribute Selection

One of the important problems in pattern recognition, machine learning, and data mining is the dimensionality reduction by attribute or feature selection. In this regard, this paper presents a feature selection method, based on interval type-2 (IT2) fuzzy-rough sets, where the features are selected by maximizing both relevance and significance of the features. By introducing the concept of lower and upper fuzzy equivalence partition matrices, the lower and upper relevance and significance of the features are defined for IT2 fuzzy approximation spaces. Different feature evaluation criteria such as dependency, relevance, and significance are presented for attribute selection task using IT2 fuzzy-rough sets. The performance of IT2 fuzzy-rough sets is compared with that of some existing feature evaluation indices including classical rough sets, neighborhood rough sets, and type-1 fuzzy-rough sets. The effectiveness of the proposed IT2 fuzzy-rough set-based attribute selection method, along with a comparison with existing feature selection and extraction methods, is demonstrated on several real-life data.

Simultaneous Feature Selection and Extraction Using Fuzzy Rough Sets

In this chapter, a novel dimensionality reduction method, based on fuzzy rough sets, is presented, which simultaneously selects attributes and extracts features using the concept of feature significance. The method is based on maximizing both relevance and significance of the reduced feature set, whereby redundancy therein is removed. The chapter also presents classical and neighborhood rough sets for computing relevance and significance of the feature set and compares their performance with that of fuzzy rough sets based on the predictive accuracy of nearest neighbor rule, support vector machine, and decision tree. The effectiveness of the proposed fuzzy rough set-based dimensionality reduction method, along with a comparison with existing attribute selection and feature extraction methods, is demonstrated on real-life data sets.

RNS based reconfigurable processor for high speed signal processing

Digital signal processing requires a large number of mathematical operations to be performed in high speed real time mode and repeatedly on a set of input data. The power requirements of the DSPs are increasing day by day along with the processing speed and chip area. Because of the limitation of power supply and space, mobile devices cannot use the general purpose processors to process digital signals, but a specialized DSP is required to provide a high performance - low cost solution. So, rather than traditional number systems, Residue Number System (RNS) is becoming attractive for their capabilities for performing addition and multiplication operation efficiently. A guideline for a novel Reconfigurable DSP Processor is proposed in this paper, based on the formulation of the Chinese remainder theorem and RNS, that can process any function using dynamic reconfigurability, which are the collections of some basic operations. Because of the Carry-free nature of RNS, this scheme can be implemented in Mobile and Wireless Computing and other fields where high speed computations are required with limited resources. The proposed architecture has been validated on Field Programmable Gate Array (FPGA).

A Scheme for Improving Bit Efficiency for Residue Number System

Residue Number System (RNS), which originates from the Chinese Remainder Theorem, offers a promising future in VLSI because of its carry-free operations in addition, subtraction and multiplication. This property of RNS is very helpful to reduce the complexity of calculation in many applications. A residue number system represents a large integer using a set of smaller integers, called residues. But the area overhead, cost and speed not only depend on this word length, but also the selection of moduli, which is a very crucial step for residue system. This parameter determines bit efficiency, area, frequency etc. In this paper we propose a new moduli set selection technique to improve bit efficiency which can be used to construct a residue system for digital signal processing environment. Subsequently, it is theoretically proved and illustrated using examples, that the proposed solution gives better results than the schemes reported in the literature. The novelty of the architecture is shown by comparison the different schemes reported in the literature.

Fuzzy-Rough MRMS Method for Relevant and Significant Attribute Selection

Feature selection refers to the problem of selecting the input attributes or features that are most effective to predict the sample categories. In this regard, a feature selection method is presented based on fuzzy-rough sets by maximizing both relevance and significance of the selected features. The paper also presents different feature evaluation criteria such as dependency, relevance, redundancy and significance for attribute selection task using fuzzy-rough sets. The performance of different rough set models is compared with that of some existing feature evaluation indices based on the predictive accuracy of nearest neighbor rule, support vector machine and decision tree. The effectiveness of fuzzy-rough set based attribute selection method, along with a comparison with existing feature evaluation indices and different rough set models, is demonstrated on two benchmark and two microarray gene expression data sets.

Identification of Co-expressed microRNAs Using Rough Hypercuboid-Based Interval Type-2 Fuzzy C-Means Algorithm

MicroRNAs are a class of small RNA molecules, which play an important regulatory role for the gene expression of animals and plants. Various studies have proved that microRNAs tend to cluster on chromosomes. In this regard, a novel clustering algorithm is proposed in this paper, integrating rough hypercuboid approach and interval type-2 fuzzy c-means. Rough hypercuboid equivalence partition matrix is used here to compute the lower approximation and boundary region implicitly for the clusters without the need of any user-specified threshold. Interval-valued fuzzifier is used to deal with the uncertainty associated with the fuzzy clustering parameters. The effectiveness of proposed method, along with a comparison with existing clustering techniques, is demonstrated on several microRNA data sets using some widely used cluster validity indices.

Clustering of microRNAs Using Rough Hypercuboid Based Fuzzy C-Means

MicroRNAs form a family of single strand RNA molecules having length of approximately 22 nucleotides that are present in all animals and plants. Various studies have revealed that microRNA tend to cluster on chromosomes. In this regard, a novel clustering algorithm is presented in this paper, integrating rough hypercuboid approach with fuzzy c-means. Using the concept of rough hypercuboid equivalence partition matrix, the lower approximation and boundary region are implicitly computed for the clusters without the need of any user specified threshold. The effectiveness of proposed technique, along with a comparison with existing clustering methods, is presented using some microRNA data sets with the help of several well known cluster validity indices.

Simultaneous Feature Selection and Extraction Using Feature Significance

Dimensionality reduction of a data set by selecting or extracting relevant and nonredundant features is an essential preprocessing step used for pattern recognition, data mining, machine learning, and multimedia indexing. Among the large amount of features present in real life data sets, only a small fraction of them is effective to represent the data set accurately. Prior to analysis of the data set, preprocessing the data to obtain a smaller set of representative features and retaining the optimal salient characteristics of the data not only decrease the processing time but also lead to more compactness of the models learned and better generalization. In this regard, a novel dimensionality reduction method is presented here that simultaneously selects and extracts features using the concept of feature significance. The method is based on maximizing both relevance and significance of the reduced feature set, whereby redundancy therein is removed. The method is generic in nature in the sense that both supervised and unsupervised feature evaluation indices can be used for simultaneously feature selection and extraction. The effectiveness of the proposed method, along with a comparison with existing feature selection and extraction methods, is demonstrated on a set of real life data sets.