A New Early Stopping Algorithm for Improving Neural Network Generalization

Depth profiling has been performed by using Auger electron spectrometry (AES) and X-ray photoelectron spectrometry (XPS) in combination with Ar-ion sputtering. The data obtained by both surface-analytical methods have been evaluated by means of factor analysis and afterwards by applying an artificial neural network or fuzzy clustering in order to determine the compositional layering of different samples such as a Cr2O3/CrN sandwich layer, tarnish layers on a nickel based alloy and on steel, and the coating of a Si3N4 ceramic powder. The applied artificial neural network was a Kohonen network. It turned out that the method of fuzzy c-means clustering was more successful than Kohonen network due to the fact that fuzzy c-means clustering starts with more input information which can be obtained from factor analysis.

Lip feature extraction is one of the most challenging tasks in the lip reading systems' performance. In this paper, a new approach for lip contour extraction based on fuzzy clustering is proposed. The algorithm employs a stochastic cost function to partition a color image into lip and non-lip regions such that the joint probability of the two regions is maximized. First, the mouth location is determined and then, lip region is preprocessed using pseudo hue transformation. Fuzzy c-means clustering is applied to each transformed image along with b components of CIELAB color space. To delete the clustered pixels around lip, an ellipse and a Gaussian mask were used. In order to show the performance of the proposed method, the pseudo hue segmentation and fuzzy c-mean clustering without preprocessing are compared. The compared methods were applied to the VidTIMIT and M2VTS databases and the results show the superiority of the proposed method in comparison with other methods.

Clustering plays a vital role in process mining as it organizes complex event logs into meaningful groups, helping to identify common patterns, outliers, and inefficiencies. This simplification enables organizations to detect bottlenecks and optimize workflows by uncovering trends and variations that might otherwise remain hidden. Fuzzy clustering addresses the challenge of overlapping behaviors, providing actionable insights for targeted improvements and enhanced operational efficiency. Nevertheless, conventional clustering algorithms for process mining focus either on activity sequences or cycle times, resulting in incomplete insights due to the neglect of temporal or structural variations. This work introduces a new fuzzy clustering methodology that incorporates both activity sequences and cycle times through a weighted distance metric. The proposed approach balances the weights of similarity in sequences as well as time variation flexibly using the parameter α, enabling clusters to represent both structural as well as performance-based process attributes. Through using fuzzy C-means clustering, the method allows cases to have multiple memberships with different membership degrees, providing flexibility regarding overlapping process behavior. An experimental evaluation using real-life event logs demonstrates the effectiveness of the method in discerning process variants. It yields superior results compared to conventional methods that account for only sequence-based clustering scenarios, as well as time-based clustering methods. The results describe the significant importance of optimizing clustering results by varying α, where a balanced weighting (α=0.5) gives more meaningful clusters. Ultimately, the framework enhances process mining by offering detailed insights for analyzing operational inefficiencies, bottlenecks, and resource allocation mismatches, providing substantial real-world benefits for industries that demand effective process improvement.

Detection of scratch defects on randomly textured surfaces remains challenging due to their unnoticeable visual features. In this paper, an algorithm for piezoelectric ceramic plate surface scratch defects based on the combination of fuzzy c-means clustering and morphological features is proposed. Foreground membership of each gray value is calculated firstly on a reference set of training images by fuzzy c-means clustering and the interpolation method, then an enhanced image is obtained by multiplying the foreground membership function and gray image. The location relationship between regions and the gradient direction of regions is extracted from the binary image of the enhanced image. Based on the morphological features, isolated non-scratched defects are filtered out and the intermittent scratches are merged. Experiments show that the algorithm can be used to detect scratch defects on the surface of a piezoelectric ceramics plate with randomly textured surfaces.

Fuzzy Clustering is applied to the problem of detecting landmines in Ground Penetrating Radar (GPR). The DARPA Backgrounds Data provides a rich source of signatures derived from a cluttered environment with a variety of sensors. One sensor used in the Backgrounds collection was the GPR developed and fielded by Geo-Centers, Inc. This GPR provides a three-dimensional array of intensity returns corresponding to a volume underneath the ground. In this paper, a novel approach to processing that GPR is described. The approach relies on computing edge direction and magnitude features in the volume and comparing them to prototypes generated using fuzzy c-means clustering. A confidence map is generated corresponding to the surface traversed by the system. The confidence map is thresholded to produce detections. Experimental results show a reduction in false alarm rates from about 40% using the standard processing method to about 4% using the three-dimensional, fuzzy clustering method.

The purpose of medical image segmentation is to divides the lesions image with a special meaning and background regional. The characteristics of medical images are generally more complex. There is often overlap between different regions, the edge of the area is vague. Research based on fuzzy clustering method for medical image segmentation, fuzzy C-means clustering and fuzzy kernel clustering methods were studied and discussed fully considered the spatial information, the single objective problem conversion into multi-objective problem. The experimental results show that the method use in medical image segmentation achieved good results.

Hazard vulnerability assessment of critical infrastructures (CIs) is crucial for ranking infrastructures based on their level of criticality, enabling the urban managers to prioritize CIs for allocating funds in the hazard mitigation/recovery process. This study aims to provide a framework for ranking CIs based on a rapid and preliminary flood vulnerability assessment by introducing a methodology for classifying CIs according to their vulnerability to riverine flooding. An indicator‐based vulnerability curve is calculated both quantitatively (using Fuzzy Logic Toolbox in MATLAB) and qualitatively (using susceptibility–exposure matrix), based on which CIs prioritization is accomplished with a focus on functional flood vulnerability considering structural/nonstructural damages. Besides, this study addresses the consequences that a damaged infrastructure may have on the rest of CIs and estimates their vulnerability given the additive impact of the surrounding failed infrastructures considering their interdependence. The methodology was applied to Berat (Albania) and Sarajevo (Bosnia‐Herzegovina) with findings compared to those of a multi‐criteria decision‐making‐based approach commonly used in CI ranking literature. The obtained results from both methods represent that roads are the most vulnerable studied infrastructure in the case of Berat, while regarding the city of Sarajevo, road infrastructures are considered the least vulnerable to riverine floods compared to bridges and schools.

Context. The problem of clustering (classification without a teacher) is often occures when processing data arrays of various natures, which is quite an interesting and integral part of artificial intelligence. To solve this problem, there are many known methods and algorithms based on the principles of the distribution density of observations in the analyzed data. However, these methods are rather complicated in software implementation and are not without drawbacks, namely: the problem of determining significant clusters in datasets of different densities, multiepoch self-learning, getting stuck in local extrema of goal functions, etc. It should be noted that the methods based on the analysis of the peaks of the data distribution density are clear in nature, therefore, to expand the capabilities of these methods, it is advisable to introduce their fuzzy modification. Objective. The aim of the work is to introduce fast fuzzy data clustering using density peaks distribution of the datasets, that can find the prototypes (centroids) of clusters that overlapping regardless of the amount of incoming data. Method. The problem of fuzzy clustering data arrays based on a hybrid method that based on the simultaneous use of a credibilistic approach to fuzzy clustering and an algorithm for finding the types of distribution density of the initial data is proposed. A feature of the proposed method is computational simplicity and high speed, due to the fact that the entire array is processed only once, that is, eliminates the need for multi-era self-learning, implemented in traditional fuzzy clustering algorithms. Results. A feature of the proposed method of fast fuzzy credibilistic clustering using of density peaks distribution is characterized by computational simplicity and high speed due to the fact that the entire array is processed only once, that is, the need for multiepoch self-learning is eliminated, which is implemented in traditional fuzzy clustering algorithms. The results of the computational experiment confirm the effectiveness of the proposed approach in clustering problems under conditions in the case when the clusters are ovelap. Conclusions. The experimental results allow us to recommend the developed method for solving the problems of automatic clustering and data classification, as quickly as possible to find the centroids of clusters. The proposed method of fast fuzzy credibilistic clustering using of density peaks distribution of dataset is intended for use in computational intelligence systems, neuro-fuzzy systems, in training artificial neural networks and in clustering problems.

This paper presents a fuzzy logic controller for the closed loop control of BLDC motor. The fuzzy logic controller is developed using a fuzzy logic toolbox in MATLAB. Also here a single current sensor technique is proposed for closed loop current control. For closed loop current control of BLDC motor, the motor phase currents are measured using current sensors. These sensors are expensive and the use of different current sensors can cause undesirable imbalance in phase currents due to differences in current sensor sensitivities. These drawbacks can be avoided by using a single current sensor placed on the DC link. Here an algorithm is presented to obtain the phase current values from the DC link current. The proposed system was simulated using MATLAB/Simulink, with PI and fuzzy logic speed controllers. The simulation result shows that, the fuzzy logic controller gives better performance than PI controller. The hardware implementation of the fuzzy logic controller is done using PIC16F877A. The hardware results show that the fuzzy logic controller gives a smooth speed control.

In fuzzy C-means (FCM) clustering, each data point belongs to a cluster to a degree specified by a membership grade. FCM partitions a collection of vectors in c fuzzy groups and finds a cluster center in each group such that the dissimilarity measure is minimized. This paper presents a training algorithm for the radial basis function (RBF) network using symmetry-based Fuzzy C-means (SFCM) clustering method which is the modified version of FCM clustering method based on point symmetry distance measure. The training algorithm which uses SFCM clustering method to train the network has a number of advantages such as faster training time, more accurate predictions and reduced network architecture compared to the standard RBF networks. The proposed training algorithm has been implemented in the RBF networks created by the newrb function of MATLAB which uses gradient based iterative method as learning strategy, therefore the new network will undergo a hybrid learning process. The networks called Symmetry-based Fuzzy C-means Clustering–Radial Basis Function Network (SFCM/RBF) has been tested against the standard RBF network and the networks called standard Fuzzy C-means Clustering (FCM)-RBF network (FCM/RBF) in forecasting. The experimental models has been tested on three real world application problems, particularly in Air pollutant problem, Biochemical Oxygen Demand (BOD) problem, and Phytoplankton problem. Keywords: Fuzzy c-means clustering; SFCM; Radial basis function network; point symmetry distance; forecasting.

Colonoscopy with biopsy is the "gold" standard for evaluating disease activity in inflammatory bowel diseases (IBD). Current research is geared toward finding non-invasive, cost-efficient methods that estimate disease activity. We aimed to develop a neural network (NN) model for the non-invasive prediction of histologic activity in IBD using routinely available clinical-biological parameters. Standard clinical-biological parameters and histologic activity from 371 ulcerative colitis (UC) and 115 Crohn's disease (CD) patient records were collected. A training set, a test set, and a validation set were used for building/validating 2 models for each disease. All models had binary output predicting the active/inactive histologic disease status. For both diseases, the first model used both clinical and biological inputs, while the second used only biological data. First UC model obtained an accuracy of 95.59% on the test set and 96.67% on the validation set. The second UC model achieved accuracies of 88.24% and 86.67% on the test and validation sets, respectively. The First CD classifier resulted in 90.48% accuracy on the test set and 91.67% on the validation set. Finally, the second CD classifier obtained an accuracy of 85.71% on the test set and 91.67% on the validation set. An accurate and non-invasive artificial intelligence system to predict histologic disease activity in IBD is designed. Our models achieved similar or better results compared to the documented performance of fecal calprotectin (the best non-invasive IBD biomarker to date). Given these favorable results, we anticipate the future utility in the clinical setting of a non-invasive disease activity prediction.

A novel kernelized total Bregman divergence-based fuzzy clustering with local information for image segmentation

A guide on the use of SVMs in pattern classification, including a rigorous performance comparison of classifiers and regressors. The book presents architectures for multiclass classification and function approximation problems, as well as evaluation criteria for classifiers and regressors. Features: Clarifies the characteristics of two-class SVMs; Discusses kernel methods for improving the generalization ability of neural networks and fuzzy systems; Contains ample illustrations and examples; Includes performance evaluation using publicly available data sets; Examines Mahalanobis kernels, empirical feature space, and the effect of model selection by cross-validation; Covers sparse SVMs, learning using privileged information, semi-supervised learning, multiple classifier systems, and multiple kernel learning; Explores incremental training based batch training and active-set training methods, and decomposition techniques for linear programming SVMs; Discusses variable selection for support vector regressors.

This paper aimed to discuss the denoising ability of magnetic resonance imaging (MRI) images based on fuzzy C-means clustering (FCM) algorithm and the influence of Butylphthalide combined with Edaravone treatment on nerve function and vascular endothelial function in patients with acute cerebral infarction (ACI). Based on FCM algorithm, Markov Random Field (MRF) model algorithm was introduced to obtain a novel algorithm (NFCM), which was compared with FCM and MRF algorithm in terms of misclassification rate (MCR) and difference of Kappa index (KI). 90 patients with ACI diagnosed in hospital from December 2018 to December 2019 were selected as subjects, who were divided into combined treatment group (conventional treatment + Edaravone + Butylphthalide) and Edaravone group (conventional treatment + Edaravone) randomly, each consisting of 45 cases. The National Institutes of Health Stroke Scale (NIHSS) score and endothelial function index level such as plasma nitric oxide (NO), human endothelin-1 (ET-1), and vascular endothelial cell growth factor (VEGF) were compared before and after treatment between the two groups. The results showed that the MCR of NFCM was evidently inferior to FCM and MRF, and the KI was notably higher relative to the other two algorithms. After treatment, the NIHSS score of the combined treatment group was (9.09 ± 1.86) points and that of Edaravone group was (14.97 ± 3.44) points, with evident difference between the two groups (P < 0.05). After treatment, the NO of the combined treatment was (54.63 ± 4.85), and that of Edaravone group was (41.54 ± 5.27), which was considerably different (P < 0.01), and the VEGF and ET-1 of combined treatment group were greatly inferior to Edaravone group (P < 0.01). It was revealed that the novel algorithm based on FCM can obtain more favorable quality and segmentation accuracy of MRI images. Moreover, Butylphthalide combined with Edaravone treatment can effectively improve nerve function, vascular endothelial function, and short-term prognosis in ACI, which was safe and worthy of clinical adoption.

PurposeThe theory of double jeopardy (DJ) is shown to hold across broad ranging geographies and physical product categories. However, there is very little research appertaining to the subject within an online environment. In particular, studies that investigate the presence of DJ and the contrasting view point to DJ, namely, that of negative double jeopardy (NDJ), are lacking. This study aims to contribute to this identified research gap and examines the presence of DJ and NDJ within a product category, utilising data from Twitter.Design/methodology/approachA total of 354,676 tweets are scraped from Twitter and their sentiment analysed and allocated into positive, negative and no-opinion clusters using fuzzy c-means clustering. The sentiment is then compared to the market share of brands within the beer product category to establish whether a DJ or NDJ effect is present.FindingsData reveal an NDJ effect with regards to original tweets (i.e. tweets which have not been retweeted). That is, when analysing tweets relating to brands within a defined beer category, the authors find that larger brands suffer by having an increased negativity amongst the larger proportion of tweets associated with them.Research limitations/implicationsThe clustering approach to analyse sentiment in Twitter data brings a new direction to analysis of such sentiment. Future consideration of different numbers of clusters may further the insights this form of analysis can bring to the DJ/NDJ phenomenon. Managerial implications discuss the uncovered practitioner’s paradox of NDJ and strategies for dealing with DJ and NDJ effects.Originality/valueThis study is the first to explore the presence of DJ and NDJ through the utilisation of sentiment analysis-derived data and fuzzy clustering. DJ and NDJ are under-explored constructs in the online environment. Typically, past research examines DJ and NDJ in separate and detached fashions. Thus, the study is of theoretical value because it outlines boundaries to the DJ and NDJ conditions. Second, this research is the first study to analyse the sentiment of consumer-authored tweets to explore DJ and NDJ effects. Finally, the current study offers valuable insight into the DJ and NDJ effects for practicing marketing managers.