The main task for a 2D spiral recognition algorithm is to learn to discriminate between data distributed on two distinct strands in the x-y plane. This problem is of critical importance since it incorporates temporal characteristics often found in real-time applications, i.e. the spiral coils with time. Previous work with this benchmark has witnessed poor results with statistical methods such as discriminant analysis and tedious procedures for better results with neural networks. This paper describes a fuzzy approach which outperforms previous work in terms of the recognition rate and the speed of recognition. The paper presents the new approach and results with the validation and test sets. The results show that it is possible to solve the spiral problem in a relatively small amount of time with the fuzzy approach (up to 100% correct classification on the validation and test set; 77.2% correct classification with crossvalidation using the leave-one-out method). Keywords: possibilisti...

This paper studies the performance of Massively Parallel Fuzzy Systems (MPFS) on the two spiral benchmark. Spiral data is contaminated with five different noise distributions. The recognition rates of the system are reported with varying levels of different types of noise. The behaviour of the system is investigated with additive, multiplicative, cumulative and non-cumulative noise. The results show that the MPFS system remains stable to different noise variations and the generalisation error remains consistently low. As the total noise in the system increases, the system witnesses a linear decrease in entropy and the generalisation error is easier to predict. The error rate is found to have two separate patterns of variation for cumulative and non-cumulative noise. Keywords: MPFS, noise distribution, possibility, recognition rate, spiral benchmark 2 1 INTRODUCTION Massively Parallel Fuzzy Systems (MPFS) were first proposed by Singh to solve pattern recognition problems in real-time...

In this paper we compare the ability of a fuzzy neural network and a classical backpropagation network to classify odour samples which were obtained by an electronic nose employing semi-conducting oxide conductometric gas sensors. Two different samples sets were analysed: first the aroma of 3 blends of commercial coffee, and secondly the headspace of 6 different tainted water samples. The two experimental data-sets provided an excellent opportunity to test the ability of a fuzzy neural network due to the high level of sensor variability often experienced with this type of sensor. Results are presented on the application of 3 layer fuzzy neural networks to electronic nose data which demonstrate a considerable improvement in performance to a common back-propagation network. 1. Introduction Artificial neural networks (ANNs) have been the subject of considerable research for over twenty years. However, it is during the last decade or so that research interest 1 Sensors and Actuators, vo...

The main purpose of this paper is to compare clustering (region growing) and gradient based techniques for detecting regions of interest in digital mammograms. These regions of interest form the basis of applying shape and texture techniques for detecting cancerous masses. In addition, the paper proposes a two stage method where gradient based techniques are applied first followed by region growing method which will yield lesser numbers of regions for analysis. For this purpose we first use histogram equalisation and fuzzy enhancement techniques to improve the quality of the images and compare their utility on our mammogram data. Image enhanced mammograms are then subjected to clustering or gradient operations (masking) for segmentation purpose. The segmented image is then analysed for estimating the regions of interest and the results are compared against previously known diagnosis of the radiologist. A total of 30 mammograms from the University of South Florida database were used whe...

The main objectives of this paper are: 1) to describe the working of a Massively Parallel Fuzzy System; 2) to test the system on a new benchmark, the three spiral data set; and 3) to describe the behaviour of the system when solving the problem. The described system is aimed at solving pattern recognition problems in real-time. Pattern recognition data are subjected to non-iterative decision making through the estimation of class membership of test data. This paper describes the performance of this system on the temporal three spiral benchmark. The task is to learn three class data which lies on three distinct spirals that coil around each other and around the origin with time. There are no linear solutions to this problem (33% recognition rate with discriminant analysis). The system under consideration classifies the training set with 100% success and recognises training data with up to 89% (without the use of rejection threshold) and 98% success (with a rejection threshold of =.8)....

The main aim of this paper is to demonstrate the performance of a possibility based classifier which implements dynamic pattern recognition. The objectives of the paper are: 1) to detail the working of a Massively Parallel Fuzzy System (MPFS) which can be used to classify two spiral data; 2) to develop and detail the dynamic pattern recognition approach; and 3) to discuss the results obtained. The results suggest that dynamic pattern recognition will perform high quality decision making for tasks involving uncertain, noisy and small amounts of data. The two spiral task can be solved with dynamic pattern recognition with more than 90% recognition success and this performance remains stable with increasingly noisy test sets. INTRODUCTION Spiral data has several interesting characteristics for validating novel neural and other pattern recognition algorithms as for example: 1) it is highly non-linear and with linear techniques a poor recognition rate is achieved; 2) it is temporal in the...

The main aim of this paper is three fold: a) to understand the working of a classifier system based on possibility distribution functions, b) to evaluate its performance against other superior methods such as fuzzy and non-fuzzy neural networks on real data, c) and finally to recommend changes for enhancing its performance. The paper explains how to construct a possibility based classifier system which is used with conventional error-estimation techniques such as crossvalidation and boot-strapping. The results were obtained on a set of electronic nose data and this performance was compared with earlier published results on the same data using fuzzy and non-fuzzy neural networks. The results show that the possibility approach is superior to the non-fuzzy approach, however, further work needs to be done. 1. Introduction Classifier systems for most pattern recognition problems need to satisfy various criteria to be considered reliable for decision making. These criteria relate not only ...

The main aim of this paper is to develop and implement the concept of incremental learning for fuzzy statistical classifiers. Such a scheme involves continuous modification of training data as learning progresses and is implemented with classifier systems that adapt to incremental information. This paper discusses the implementation of the above approach using a real-time fuzzy classifier system. The recognition performance of this approach on the three spiral benchmark is compared with conventional static training. The paper also discusses the generalisation performance of the system in the context of recognising noisy spiral data. 2 1. Introduction Fuzzy pattern recognition refers to the use of fuzzy mathematical approaches to pattern recognition. The development of fuzzy mathematical approaches for classification purposes is important in several areas including speech analysis, image processing and fuzzy control [21]. This area of research is important to advance the performance o...