Showing papers on "Fuzzy clustering published in 1994"

Generation of Fuzzy Rules by Mountain Clustering

Abstract: We develop, based upon the mountain clustering method, a procedure for learning fuzzy systems models from data. First we discuss the mountain clustering method. We then show how it could be used to obtain the structure of fuzzy systems models. The initial estimates of this model are obtained from the cluster centers. We then use a back propagation algorithm to tune the model.

MRI segmentation using fuzzy clustering techniques

Abstract: The authors' main contribution is to build upon their earlier efforts by expanding the tissue model concept to cover a brain volume. Furthermore, processing time is reduced and accuracy is enhanced by the use of knowledge propagation, where information derived from one slice is made available to succeeding slices as additional knowledge. The system is organized as follows. Each MR slice is initially segmented by an unsupervised fuzzy c-means clustering algorithm. Next, an expert system uses model-based recognition techniques to locate a landmark, called a focus-of attention tissue. Qualitative models of slices of brain tissue are defined and matched with their instances from imaged slices. If a significant deformation is detected in a tissue, the slice is classified to be abnormal and volume processing halts. Otherwise, the expert system locates the next focus-of-attention tissue, based on a hierarchy of expected tissues. This process is repeated until either a slice is classified as abnormal or all tissues of the slice are labeled. If the slice is determined to be abnormal, the entire volume is also considered abnormal and processing halts. Otherwise, the system will proceed to the next slice and repeat the classification steps until all slices that comprise the volume are processed. A rule-based expert system tool, CLIPS, is used to organize the system. Low level modules for image processing and high level modules for image analysis, all written in the C language, are called as actions from the right hand sides of the rules. The system described here is an attempt to provide completely automatic segmentation and labeling of normal volunteer brains. The absolute accuracy of the segmentations has not yet been rigorously established. The relative accuracy appears acceptable. Efforts have been made to segment an entire volume (rather than merging a set of segmented slices) using supervised pattern recognition techniques or unsupervised fuzzy clustering. However, there is sometimes enough data nonuniformity between slices to prevent satisfactory segmentation. >

Detection and separation of ring-shaped clusters using fuzzy clustering

Abstract: A new fuzzy clustering algorithm, designed to detect and characterize ring-shaped clusters and combinations of ring-shaped and compact spherical clusters, has been developed. This FKR algorithm includes automatic search for proper initial conditions in the two cases of concentric and excentric (intersected) combinations of clusters. Validity criteria based on total fuzzy area and fuzzy density are used to estimate the optimal number of substructures in the data set. The FKR algorithm has been tested on a variety of simulated combinations of ring-shaped and compact spherical clusters, and its performance proved to be very good, both in identifying the input shapes and in recovering the input parameters. Application of the FKR algorithm to an MRI image of the heart's left ventricle was used to investigate the possibility of using this algorithm as an aid in image processing. >

Genetic algorithm guided clustering

Abstract: Genetic algorithms provide an approach to optimization. Unsupervised clustering algorithms attempt to optimize the placement of like objects into homogeneous classes or clusters. We describe an approach to using genetic algorithms to optimize the clusters created during unsupervised clustering. Hard partitions of the feature space are the members of the population. They evolve into better partitions based upon the fitness function which is a version of the hard c-means optimization function. The methods of crossover and mutation are described. An example of the clustering performance of this approach is shown with the Iris data. The genetic guided clustering is shown to outperform hard c-means on the Iris data in terms of the number of patterns which are correctly placed into a partition whose majority class is the same as the assigned pattern. >

A least biased fuzzy clustering method

Abstract: A new operational definition of cluster is proposed, and a fuzzy clustering algorithm with minimal biases is formulated by making use of the maximum entropy principle to maximize the entropy of the centroids with respect to the data points (clustering entropy). The authors make no assumptions on the number of clusters or their initial positions. For each value of an adimensional scale parameter /spl beta/', the clustering algorithm makes each data point iterate towards one of the cluster's centroids, so that both hard and fuzzy partitions are obtained. Since the clustering algorithm can make a multiscale analysis of the given data set one can obtain both hierarchy and partitioning type clustering. The relative stability with respect to /spl beta/' of each cluster structure is defined as the measurement of cluster validity. The authors determine the specific value of /spl beta/' which corresponds to the optimal positions of cluster centroids by minimizing the entropy of the data points with respect to the centroids (clustered entropy). Examples are given to show how this least biased method succeeds in getting perceptually correct clustering results. >

Algorithms for clustering molecular dynamics configurations

Abstract: Two traditional clustering algorithms are applied to configurations from a long molecular dynamics trajectory and compared using two sets of test data. First, a subset of atoms was chosen to present conformations which naturally fall into a number of clusters. Second, a subset of atoms was selected to span a relatively continuous region of conformational space rather than form discrete conformational classes. Of the two algorithms used, the single linkage method is inappropriate for this kind of data. The divisive hierarchical method, based on minimizing the difference between cluster centroids and extrema, is successful but also prone to imposing clustering hierarchy where none can be justified. 0 1994 by John Wiley & Sons, Inc.

Optimization of fuzzy clustering criteria using genetic algorithms

Abstract: This paper introduces a general approach based on genetic algorithms for optimizing a broad class of clustering criteria. The standard approach for optimizing these criteria has been to alternate optimizations between the variables which represent fuzzy memberships of the data to various clusters, and those prototype variables which determine the geometry of the clusters. The approach suggested here first re-parameterizes the criteria into functions of the prototype variables alone. The prototype variables are then coded as binary strings so that genetic algorithms can be applied. An overview of the approach and two simple numerical examples are given. >

Identification of fuzzy prediction models through hyperellipsoidal clustering

Abstract: To build a fuzzy model, as proposed by Takagi and Sugeno (1985), the authors emphasize an interactive approach in which knowledge or intuition can play an important role It is impossible in principle, due to the nature of the data, to specify a criterion and procedure to obtain an ideal fuzzy model The main subject of fuzzy modeling is how to analyze data in order to summarize it to a certain extent so that one can judge the quality of a model by intuition The main proposal in this paper is a clustering technique which takes into account both continuity and linearity of the data distribution The authors call this technique the hyperellipsoidal clustering method, which assists modelers in finding fuzzy subsets suitable for building a fuzzy model The authors deal with other problems in fuzzy modeling as well, such as the effect of data standardization, the selection of conditional and explanatory variables, the shape of a membership function and its tuning problem, the manner of evaluating weights of rules, and the simulation technique for verifying a fuzzy model >

MECA: maximum entropy clustering algorithm

Abstract: This paper presents a new approach to fuzzy clustering, which provides the basis for the development of the maximum entropy clustering algorithm (MECA). The derivation of the proposed algorithm is based on an objective function incorporating a measure of the entropy of the membership functions and a measure of the distortion between the prototypes and the feature vectors. This formulation allows the gradual transition from a maximum uncertainty or minimum selectivity phase to a minimum uncertainty or maximum selectivity phase during the clustering process. Such a transition is achieved by controlling the relative effect of the maximization of the membership entropy and the minimization of the distortion between the prototypes and the feature vectors. The IRIS data set provides the basis for evaluating the proposed algorithms and comparing their performance with that of competing techniques. >

Generation of membership functions via possibilistic clustering

Abstract: Possibilistic clustering has been introduced recently to overcome some of the limitations imposed by the constraint used in the fuzzy c-means algorithm. It was shown that possibilistic memberships correspond more closely to the notion of "typicality". In this paper, we explore certain interesting properties of possibilistic clustering, In particular, we show that possibilistic clustering can be successfully used to solve two important problems that arise while using fuzzy set theory: i) determination of membership functions, and ii) determination of the number of clusters. >

Performance of Fuzzy ART neural network for group technology cell formation

Abstract: This study investigates the performance of Fuzzy ART neural network for grouping parts and machines, as part of the design of cellular manufacturing systems. Fuzzy ART is compared with ART1 neural network and a modification to ART1, along with direct clustering analysis (DCA) and rank order clustering (ROC2) algorithms. A series of replicated clustering experiments were performed, and the efficiency and consistency with which clusters were identified were examined, using large data sets of differing sizes and degrees of imperfection. The performance measures included the recovery ratio of bond energy and execution times, It is shown that Fuzzy ART neural network results in better and more consistent identification of block diagonal structures than ART1, a recent modification to ART1, as well as DCA and ROC2. The execution times were found to be more than those of ART1 and modified ART1, but they were still superior to traditional algorithms for large data sets.

Multidimensional Scaling and Data Clustering

Abstract: Visualizing and structuring pairwise dissimilarity data are difficult combinatorial optimization problems known as multidimensional scaling or pairwise data clustering. Algorithms for embedding dissimilarity data set in a Euclidian space, for clustering these data and for actively selecting data to support the clustering process are discussed in the maximum entropy framework. Active data selection provides a strategy to discover structure in a data set efficiently with partially unknown data.

A hybrid artificial neural network-dynamic programming approach for feeder capacitor scheduling

Abstract: A hybrid artificial neural network (ANN) dynamic programming (DP) method for optimal feeder capacitor scheduling is presented in this paper. To overcome the time-consuming problem of full dynamic programming method, a strategy of ANN assisted partial DP is proposed. In this method, the DP procedures are performed on historical load data offline. The results are managed and valuable knowledge is extracted by using cluster algorithms. By the assistance of the extracted knowledge, a partial DP of reduced size is then performed online to give the optimal schedule for the forecasted load. Two types of clustering algorithms, hard clustering by Euclidean algorithm and soft clustering by an unsupervised learning neural network, are studied and compared in the paper. The effectiveness of the proposed algorithm is demonstrated by a typical feeder in Taipei City with its 365 days' load records. It is found that execution time of scheduling is highly reduced, while the cost is almost the same as the optimal one derived from full DP. >

A fuzzy clustering method for the identification of fuzzy models for dynamic systems

Abstract: Fuzzy modeling is an important topic in fuzzy sets theory and applications. One particular fuzzy model structure, which can be used effectively to describe the behaviour of complex nonlinear systems, has been given by Takagi and Sugeno (1985). By means of a fuzzy clustering method, a new approach to the identification of this kind of fuzzy model is proposed, which integrates the structure and parameter identification steps, and/or the premise and consequence identification. >

Myoelectric signal recognition using fuzzy clustering and artificial neural networks in real time

Abstract: Application of EMG-controlled functional neuromuscular stimulation to a denervated muscle depends largely on the successful discrimination of the myoelectric signal (MES) by which the subject desires to execute control over the impeded movement. This can be achieved by an adaptive and flexible interface that is robust regardless of electrode location, strength of remaining muscle activity or even personal conditions. A real-time application of an artificial neural network that can accurate recognize the MES signature is proposed in this paper. MES features are first extracted through Fourier analysis and clustered using the fuzzy c-means algorithm. Data obtained by this unsupervised learning technique are then automatically targeted and presented to a multilayer perceptron type neural network. For real-time operation, a digital signal processor operates over the resulting set of weights and maps the incoming signal to the stimulus control domain. Results show a highly accurate discrimination of the control signal over interference patterns. >

An adaptive integrated fuzzy clustering model for pattern recognition

Abstract: The extension of neural net based crisp clustering algorithms to fuzzy clustering algorithms has been addressed by many researchers in recent years. However, such neuro-fuzzy clustering algorithms developed so far suffer from restrictions in identifying the actual decision boundaries among clusters with overlapping regions. These restrictions are induced by the choice of the similarity measure and representation of the clusters. An integrated adaptive fuzzy clustering (IAFC) algorithm is presented to generate improved decision boundaries by introducing a new similarity measure and by integrating the advantages of the fuzzy optimization constraint of fuzzy c-means (FCM), the control structure of adaptive resonance theory (ART-1), and a fuzzified Kohonen-type learning rule. The effect of the new similarity measure in finding nonlinear decision boundaries among closely located cluster centroids is demonstrated with computer generated data. We use the IRIS data set and a subset of the tethered satellite system simulation data set to compare the convergence rate and misclassifications resulting from IAFC algorithm with other clustering algorithms.

Genetic fuzzy clustering

Abstract: This paper describes a genetic guided fuzzy clustering algorithm. The fuzzy-c-means functional J/sub m/ is used as the fitness function. In two domains the approach is shown to avoid some higher values of J/sub m/ to which the fuzzy-c-means algorithm will converge under some initializations. Hence, the genetic guided approach shows promise as a clustering tool. >

Color image segmentation using fuzzy clustering and supervised learning

Abstract: We propose a technique for the segmentation of color map images by means of an algorithm based on fuzzy clustering and prototype optimization. Its purpose is to facilitate the extraction of lines and characters from a wide variety of geographical map images. In this method, segmentation is considered to be a process of pixel classification. The fuzzy c-means clustering algorithm is applied to a number of training areas taken from a selection of different color map images. Prototypes, generated from the clustered pixels, that satisfy a set of validation criteria are then optimized using a neural network with supervised learning. The image is segmented using the optimized prototypes according to the nearest neighbor rule. The method has been verified to work efficiently with real geographical map data.

Performance of Selected Part‐Machine Grouping Techniques for Data Sets of Wide Ranging Sizes and Imperfection

Abstract: This study addresses the part-machine grouping problem in group technology, and evaluates die performance of several cell formation methods for a wide range of data set sizes. Algorithms belonging to four classes are evaluated: (1) array-based methods: bond energy algorithm (BEA), direct clustering analysis (DCA) and improved rank order clustering algorithm (ROC2); (2) non-hierarchical clustering method: ZODIAC; (3) augmented machine matrix methods: augmented p-median method (APM) and augmented linear clustering algorithm (ALC); and (4) neural network algorithms: ART1 and variants: ART1/KS, ART1/KSC, and Fuzzy ART. The experimental design is based on a mixture-model approach, utilizing replicated clustering. The performance measures include Rand Index and bond energy recovery ratio, as well as computational requirements for various algorithms. Experimental factors include problem size, degree of data imperfection, and algorithm tested. The results show that, among the algorithms applicable for large, industry-size data sets, ALC and neural networks are superior to ZODIAC, which in turn is generally superior to array-based methods of ROC2 and DCA.

A general framework for vertex orderings, with applications to netlist clustering

Abstract: We present a general framework for the construction of vertex orderings for netlist clustering. Our WINDOW algorithm constructs an ordering by iteratively adding the vertex with highest attraction to the existing ordering. Variant choices for the attraction function allow our framework to subsume many graph traversals and clustering objectives from the literature. The DP-RP method of [3] is then applied to optimally split the ordering into a k-way clustering. Our approach is adaptable to use-specified cluster size constraints. Experimental results for clustering and multi-way partitioning are encouraging.

Robust shape detection using fuzzy clustering: practical applications

Abstract: Problem of accurate shape detection in practical applications is considered. The work is motivated by several experimental studies in granular flow research, where in general, the objects imaged are spherical particles. A new algorithm called the divide and conquer noise fuzzy c -shells clustering (D&C-NFCS) is proposed, with these particular applications in mind, for detecting circles and ellipses in noisy images without requiring prior knowledge of the exact number of clusters. This unsupervised algorithm uses Hough transform (HT) based methods to provide very rough initial estimates of the cluster prototypes for use in the fuzzy c -shells (FCS) type algorithms. The results of HT are also used to segment the raw data so that the FCS algorithm can be applied to detect one cluster at a time. The concept of recently introduced noise clustering algorithm is also used to make the algorithm robust against noise. Results of this algorithm for several practical examples from granular flow experiments are shown demonstrating high speed and accuracy. When compared with the methods based on HT alone, this approach results in a significant improvement in detection accuracy. Cluster validity issues are also discussed.

Fuzzy clustering with genetic search

Abstract: Pattern classification task consists of clustering the training samples into known classes and using these clusters to classify new samples. Clustering is done by finding an appropriate set of ellipsoids for enclosing each of the classes. To obtain fuzzy clustering, membership values are assigned to samples against ellipsoids of all classes and these values are defuzzified for final classification. During the clustering phase, a variant of genetic algorithms, which allows variable-length genotypes, is employed in searching for the set of ellipsoids for all the classes. In particular, the number of clusters is not assumed to be known a priori, and is, in effect, determined by the genetic search dynamically. The evaluation function drives the search towards a set of ellipsoids which maximizes the correctness of classification of the training samples while having minimum total volume. >