Search

The mechanism of L2-norm loss can be explained from the perspective of maximizing margin and minimizing margin variance, which is equivalent to the idea of margin distribution. Recent studies suggest that the margin distribution is more vital for the generalization performance of models. Inspired by the idea, a nonparallel support vector machine (NPSVM) with L2-norm loss is proposed in this paper, termed as L2-NPSVM. We first reconstruct a L2-norm loss for NPSVM. The L2-norm loss can make full use of the margin distribution information in the training samples. Based on the L2-norm loss, we formulate the L2-NPSVM algorithm. L2-NPSVM is a more reliable training model with stronger generalization ability. Since L2-norm loss is used to measure the empirical risk in the objective function, L2-NPSVM can also reduce the impact of outliers. Furthermore, due to the L2-norm loss, the dual coordinate descent (DCD) method can be applied to linear and nonlinear L2-NPSVM. Under the premise of high accuracy, DCD can accelerate the solving process. At last, we conduct a series of comparative experiments on UCI datasets, noise-added UCI datasets and real-world datasets, which demonstrate the feasibility and effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

A speaker identification framework that combines both high- and low-level features, for state-of-the-art variability analysis and classification is proposed in this paper. The proposed framework introduces a workable solution that targets robust speaker variability classification using speech samples recorded in suboptimal conditions. A translated Ibibio (New Benue-Congo, Nigeria) version of "The Tiger and the Mouse"—a prosodically balanced corpus that demonstrates the prosody of read-aloud English was used in this study with speech samples obtained from 50 participants (25 males and 25 females). Identity-vectors (i-vectors) or low dimensional signal patterns were extracted and used as baselines for investigating speakers' variability patterns across various classifiers (Decision Tree: DT, Support Vector Machine: SVM, k-Nearest Neighbour: k-NN, and Deep Neural Network: DNN) and kernels. Treatment of the baselines with high level features (speech duration, F0, intensity) was also experimented for word, syllable, and phoneme units. Results obtained revealed that DTs and some variants of SVM gave high classification accuracies (above 70%). Hence, the hypothesis of universal Gaussianity appears inexact, as the linear predictor that is optimal in the mean square error sense may not hold for Ibibio. Further treatments of the baselines with Linear Discriminant Analysis (LDA) and Cosine Distant Scoring (CDS) yielded very poor classification results, except for the k-NN and Gaussian SVM classifiers which performed well for the LDA treated baselines. [ABSTRACT FROM AUTHOR]

Color images usually have large storage sizes as they are composed of three planes in the raw image format represented with the red, green, and blue components. Traditional color image compression schemes can be used to save the storage size of the color images. Unfortunately, most of these schemes are lossy in nature, which affects the details of color images. This paper presents a different treatment to the color image storage problem depending on the original color image formation process. In the color image formation process, not all the red, green, and blue components of the color images are acquired, simultaneously. Only, one component at each pixel position is acquired and Color Filter Array (CFA) interpolation is used to estimate the other two components using interpolation algorithms like Minimized-Laplacian Residual Interpolation (MLRI) and Linear Interpolation with Laplacian Second Order Correction (LILSOC). We adopt a similar strategy in this paper for reducing the storage sizes of color images by 66.67% of their original sizes. The sensitivity of the pattern recognition process to the proposed color image storage and interpolation strategy is studied in this paper. A cepstral feature extraction algorithm is adopted in this paper for extracting features from the interpolated patterns for further classification. Moreover, two types of classifiers are considered and compared in this paper for the pattern recognition: Artificial Neural Networks (ANNs), and Support Vector Machines (SVMs). Simulation results reveal the success of the proposed strategy for color image storage and interpolation in obtaining high-quality color images in addition to the high Recognition Rates (RR) of color patterns after interpolation. This success encourages the use of the proposed color image storage strategy in storing large volumes of color databases. [ABSTRACT FROM AUTHOR]

QR codes are often placed on complex backgrounds and under unstable illumination conditions, which in fact renders their localization and decoding quite challenging. In this paper, we propose a robust system devised to ensure two main tasks: firstly, extracting QR codes from captured images regardless of their quality. Secondly, enhance the extracted QR codes textures so that they can be decoded correctly. A QR code is principally localized through three finder patterns (FPs), which are placed in its three corners, as well as a set of alignment patterns (APs). FP and AP are characterized by specific textures (alternative black and white pixel sequences), defined respectively by the following ratios: 1:1:3:1:1 and 1:1:1:1:1. The proposed system starts by segmenting the captured image using an achromatic filter. This is followed by horizontal and vertical scans, aiming at extracting the valid horizontal and vertical segments satisfying the aforementioned ratios. Afterwards, the valid segment intersection is computed to localize candidate areas containing FPs and APs. The rough extracted patterns are filtered based on Multiclass Support Vector Machine classifier combined with the histogram of Oriented Gradients. The retained patterns are then transmitted to a pattern closeness measurement algorithm to conveniently localize QR code positions. In order to evaluate the robustness and weakness of the proposed approach, a comparative study has been carried out. In this study, we compared our approach against the most relevant competitive QR code extractors, such as Zxing, Zbar, Quirc, Leadtools and Dynamsoft. Our algorithm outperforms the evoked extractors with a precision rate of 97,3% and 96% under motion blur and JPEG compression respectively. Furthermore, the execution time analysis proved that our algorithm is not computationally expensive with a maximum running time of 1092 ms at a high resolution image and high number of QR codes per image.. [ABSTRACT FROM AUTHOR]

Acquisition, classification, and analysis of seismic data are crucial tasks in volcano monitoring. The large number of seismic signals that are continuously acquired during the first monitoring stage poses a huge challenge for the human experts that must classify and analyze them. Several automatic classification systems have been proposed in the literature to alleviate such an overwhelming workload, each one characterized by different levels of accuracy, computational complexity, and interpretability. Considering this last perspective, which represents one of the recent key issues in geoscience, it is possible to find many accurate methods (in terms of classification accuracy) which however represent black boxes, not permitting a clear interpretation. On the other hand, there are other approaches, such as those based on support vector machines (SVM), random forests (RF), and K-nearest neighbor (KNN), which permit the interpretation of results, rules, and models at different levels. Among these last techniques, KNN approaches for volcanic signal classification typically do not achieve the satisfactory classification results obtained with RF and SVM. One possible reason is that in this context, the KNN rule has usually been applied in its basic version, not exploiting the different advanced KNN variants that have been introduced in recent years. This paper takes one step along this direction, investigating the suitability of a number of advanced versions of the KNN rule for the problem of classifying seismic-volcanic signals. The usefulness of these rules, in comparison with the original KNN rule as well as other interpretable classifiers, is evaluated within a real-world scenario involving a five-class dataset of seismic signals acquired at the Nevado del Ruiz volcano, Colombia. The results show that the classification accuracy of basic KNN is largely improved by these advanced variants, even surpassing that obtained with other classifiers like RF and SVM. [ABSTRACT FROM AUTHOR]

Research on body-worn sensors has shown how they can be used for the detection of falls in the elderly, which is a relevant health problem. However, most systems are trained with simulated falls, which differ from those of the target population. In this paper, we tackle the problem of fall detection using a combination of novelty detectors. A novelty detector can be trained only with activities of daily life (ADL), which are true movements recorded in real life. In addition, they allow adapting the system to new users, by recording new movements and retraining the system. The combination of several detectors and features enhances performance. The proposed approach has been compared with a traditional supervised algorithm, a support vector machine, which is trained with both falls and ADL. The combination of novelty detectors shows better performance in a typical cross-validation test and in an experiment that mimics the effect of personalizing the classifiers. The results indicate that it is possible to build a reliable fall detector based only on ADL. [ABSTRACT FROM AUTHOR]

Statistical process control (SPC) has proven to be an effective tool for measuring, controlling, and improving a process through the application of statistical procedures. The most valuable SPC tool is control charts, however, if data assumptions are not met, alterations of the statistical performance of traditional control charts may be expected. In practice, the independence assumption is commonly violated in many discrete and continuous industrial processes. Recently, machine learning has been applied in the field of control chart pattern recognition (CCPR) in order to detect and identify abnormal patterns associated with assignable causes. In several instances, the performance of SVMs has been proven superior to other machine learning algorithms. In this paper, a SVM-GA-based monitoring system is presented for fault identification in AR(1) auto-correlated processes. A two stages classification scheme is presented. The initial classification stage is centered on abnormal pattern identification. The integration of a genetic algorithm in this stage boosts the performance of the SVM by simultaneously optimizing the input feature vector, the classifiers' hyper-parameters, and the window size. A second stage provides additional information on the detected abnormal pattern in order to facilitate the fault diagnosis. The results showed a good accuracy performance on both classification stages. The proposed monitoring system yielded better classification accuracies for positive autocorrelation levels than it did for their negative counterparts. The worst classification accuracies were obtained when autocorrelation was not present. [ABSTRACT FROM AUTHOR]

Many variations of local binary pattern (LBP) were proposed to enhance its performance, including uniform local binary pattern (ULBP), center-symmetric local binary patterns (CS-LBP), center symmetric local ternary patterns (CS-LTP), center symmetric local multilevel pattern (CS-LMP), etc. In this paper, the accuracies of LBP technique and its variations are enhanced using four different sizes of a sliding window approach. This approach is used for investigating whether the features extracted by LBP are significant enough or its versions are needed as well. Five LBP-based techniques have been used including LBP, CS-LBP, CS-LTP, CS-LMP, and U2LBP. They have been applied to an Arabic digit image dataset called MAHDBase. Support vector machine (SVM) and random forests are utilized as classifiers. The experimental results show that the obtained accuracies have been improved by 19.56%, 21.43%, 5.63%, 6.51% and 5.62% for CS-LBP, CS-LMP, U2LBP, CS-LTP, and LBP, respectively, when the sliding window approach has been applied and SVM with linear kernel has been used as a classifier. Moreover, the results show that there is no need to use LBP variations to enhance the accuracy if the sliding window is applied because the highest accuracy has been acquired using LBP. At the end, the accuracy of proposed systems has been compared against other state-of-the-art LBP-based techniques showing the significance of the proposed systems. [ABSTRACT FROM AUTHOR]

Implementation of adaptive modulation techniques based on the channel conditions play a key role in future wireless applications. Due to these adaptive techniques, the performance of the receiver depends on the ability of Blind Modulation Recognition (BMR). This paper investigates the performance of different Pattern Recognition Classifiers (PRC) in modulation recognition under multipath fading channels. Investigations carried out on various non-parametric supervised classifiers which include Decision Tree (DT), K-Nearest Neighbor (KNN), Support Vector Machine (SVM) and Ensemble Classifiers. To train the classifiers initially a set of 39 features is extracted and then a set of 11 noise robust features is chosen based on investigations through rough set theory. Performance of proposed PRC's is compared with other classifiers stated in literature to prove their superiority in modulation classification under fading conditions. [ABSTRACT FROM AUTHOR]

In statistical process monitoring, data mining algorithms are applied for control chart pattern recognition (CCPR) not only to detect but also to identify abnormal patterns associated with assignable causes. Recently, the use of support vector machines (SVMs) has achieved remarkable results in statistical process control applications due to its excellent generalization performance. Although there is a lot of research that highlights the superiority of support vector machine over other algorithms used in systems based on data mining control, there are no studies that analyze the design elements of a SVM classifier. Consequently, a comprehensive review, classification and an analysis of information found in the literature are presented in this paper regarding to the SVM classifier. The aim of this research is to provide researchers a starting point to potentiate the performance of the SVM classifier for assuring the best possible classification and improving the detection efficiency. Sixty-one research articles from 2001 to 2015 are critically analyzed based on their methodology following a classification scheme derived from the support vector classification framework. The analysis showed that the feature extraction and selection play a crucial role on the performance of classifiers. Studies revealed that using extracted features or a combination of raw data and feature extraction achieves better classification performances than using only raw data as an input. On the other hand, there is an ample gap to extend the research on internal structure configurations under various types of kernels, multi-classification, and ensemble approaches. Finally, this paper evidences that the application of nature inspired algorithms for kernel parameter selection in auto-correlated SVM-based process monitoring systems remains unexplored. [ABSTRACT FROM AUTHOR]

The systems based on image processing have numerous applications in the domain of motion control of robots and autonomous vehicles. The current paper is oriented to the solution of the problem that precedes the implementation of automatic avoidance of the on-road obstacles-how to detect them, to track in the sequence of images, and to recognize which of them are stationary, incoming, or outgoing from the camera. The overall algorithm of obstacle classification presented in this paper consists of three basic phases: (1) image segmentation in order to extract the pixels belonging to the image of a road and the objects over it; (2) extraction of characteristic points inside the area of the obstacle, their description and tracking in following frames; and (3) estimation of distances between the camera, the obstacles and their rates of change (relative velocities). The verifications of particular steps of the proposed algorithm are illustrated using real road-traffic images, while the overall algorithm is tested using both synthesized sequences of images and the ones acquired in real driving. [ABSTRACT FROM AUTHOR]

Based on the recently proposed twin support vector machine and twin bounded support vector machine, in this paper, we propose a novel twin support vector machine (NTSVM) for binary classification problems. The significance of our proposed NTSVM is that the objective function is changed in the spirit of regression, such that hyperplanes separate as much as possible. In addition, the successive overrelaxation technique is used to solve quadratic programming problems to speed up the training process. Experimental results obtained on several artificial and UCI benchmark datasets show the feasibility and effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

Medical decision support systems have been a core of intense research for years. The ongoing study shows that artificial intelligence has been accustomed to probe risk factors for hypertension. Factors, like health-damaging personal behaviors and changes in lifestyle and environment, are major contributors to chronic diseases. The goal of this research was to forecast the risk of developing hypertension by revealing hidden patterns in medical datasets. Quality of the data is the key to enhance the performance of learning model. But most healthcare data suffer from class imbalance problem, which induce the need for an intelligent model which can learn from such grimy data. This paper incorporates a novel approach by combing learning model and rule-based mining to offer decision support. Typically, the proposed work comprises two main implications. First suggests an intelligent learning model using boosting-based support vector machine to diagnose and expose multi-class categories in the imbalanced datasets. Finally, the enhanced predictive model is built upon the classification solution which will portray the innate data similarities. An intelligent fuzzy-based approach was employed to recognize frequent behavioral patterns. Based on these rules, valid decisions could be made to prevent hypertension. The suggested enhanced model is evaluated using a real-time hypertension dataset obtained through primary health centers. With the combination of ensemble strategies, the proposed intelligent learning model attains high classification accuracy for the imbalanced dataset above the traditional model. Thus, the efficient integration of personalized behavior with health data could provide a better understanding regarding patient health. In future this can serve as an eye toward personalized medicine. [ABSTRACT FROM AUTHOR]

A new approach for face recognition, based on kernel principal component analysis (KPCA) and support vector machines (SVMs), is presented to improve the recognition performance of the method based on principal component analysis (PCA). This method can simultaneously be applied to solve both the over-fitting problem and the small sample problem. The KPCA method is performed on every facial image of the training set to get the core facial features of the training samples. To ensure that the loss of the image information will be as less as possible, the facial data of high-dimensional feature space is projected into low-dimensional space, and then the SVM face recognition model is established to identify the low-dimensional space facial data. Our experimental results demonstrate that the approach proposed in this paper is efficient, and the recognition accuracy of the proposed method reaches 95.4 %. [ABSTRACT FROM AUTHOR]

New channels, new carriers continue to produce new data types, how to deal with complex and volatile multi-source data under IoT big data environment has become an important issue. Information perception is the basic content of the development of things, but also an important means to achieve information perception. Information interaction is the basic condition for the realization of Internet of things, but also the basic technology of interconnection. The information sensing of the Internet of Things is mainly the source of information provided by it, which is also the application base of the Internet of Things itself. To effectively process the IoT data, this paper proposes the multimodal intelligent data analyzing paradigm for Internet of things systems. Our proposed multiple game theory based data analyzing paradigm will guarantee the optimal combination of the mentioned model to construct the effect and robust analyzing pattern. We construct the optimal combination of Bayesian model, Decision Tree, Random Forest, Logistic Regression, SVM, Adaboost, Markov and Neural Network under the multiple game framework. Several new ideas are proposed to enhance the traditional Nash Equilibrium. Performance of the proposed method is validated through the simulation compared with the other state-of-the-art algorithms. [ABSTRACT FROM AUTHOR]

In this paper, we propose an anomaly detection system of machines using a hybrid learning mechanism that combines two kinds of machine learning approaches, namely unsupervised and non-parametric learning. To do so, we used vibration data, which is known to be suitable for anomaly detection in machines during operation. Furthermore, in order to take into account various characteristics of abnormal data such as scarcity and diversity, we propose a novel method that can detect anomalous behaviors using normal patterns instead of abnormal patterns from the machines. That is, we first perform a machine learning of the normal patterns of the machines during operation, and if any of the operation patterns deviates from the normal pattern, we identify that pattern as abnormal. A key characteristic of our system is that it does not use any prior information such as predefined data labels or data distributions to learn the normal operation patterns. To demonstrate the superiority of our system, we constructed a test bed consisting of a washing machine and a 3-axis accelerometer. We also demonstrated that our system can improve the accuracy of anomaly detection for the machines compared to other approaches. [ABSTRACT FROM AUTHOR]

The asymmetry of different misclassification costs is a common problem in many realistic applications. However, most of the traditional classifiers pursue high recognition accuracy, assuming that different misclassification errors bring uniform cost. This paper proposes two cost-sensitive models based on support vector data description (SVDD) to minimize classification costs while maximize classification accuracy. The one-class classifier SVDD is extended to two two-class models. The cost information is incorporated to pursue tradeoff generalization performances between different classes in order to minimize the misclassification costs. Cost information is also considered to build the decision rules. The solutions of the optimization problems of the proposed two models are formulated according to sequential minimal optimization (SMO) algorithm. However, SMO needs to check all the samples to select the working set in each iteration, which is very time consuming. Considering that only the support vectors are needed to describe the boundaries, a sample selection approach is proposed to speed up the training time and reduce the storage requirement by selecting edge and overlapping samples, and overcome the local overlearning by remove outliers. Experimental results on synthetic and public datasets demonstrate the effectiveness and efficiency of the proposed methods. [ABSTRACT FROM AUTHOR]

Nowadays, millions of people are affected by heart diseases worldwide, whereas a considerable amount of them could be aided through an electrocardiogram (ECG) trace analysis, which involves the study of arrhythmia impacts on electrocardiogram patterns. In this work, we carried out the task of automatic arrhythmia detection in ECG patterns by means of supervised machine learning techniques, being the main contribution of this paper to introduce the optimum-path forest (OPF) classifier to this context. We compared six distance metrics, six feature extraction algorithms and three classifiers in two variations of the same dataset, being the performance of the techniques compared in terms of effectiveness and efficiency. Although OPF revealed a higher skill on generalizing data, the support vector machines (SVM)-based classifier presented the highest accuracy. However, OPF shown to be more efficient than SVM in terms of the computational time for both training and test phases. [ABSTRACT FROM AUTHOR]

Accurate control chart patterns recognition (CCPR) plays an essential role in the implementation of control charts. However, it is a challenging problem since nonrandom control chart patterns (CCPs) are normally distorted by 'common process variations'. In this paper, a novel method of CCPR by integrating fuzzy support vector machine (SVM) with hybrid kernel function and genetic algorithm (GA) is proposed. Firstly, two shape features and two statistical features that do not depend on the distribution parameters and number of samples are presented to explicitly describe the characteristics of CCPs. Then, a novel multiclass method based on fuzzy SVM with a hybrid kernel function is proposed. In this method, the influence of outliers on classification accuracy of SVM-based classifiers is weakened by assigning a degree of membership for every training sample. Meanwhile, a hybrid kernel function combining Gaussian kernel and polynomial kernel is adopted to further enhance the generalization ability of the classifiers. To solve the issue of features selection and parameters optimization, GA is used to simultaneously optimize the input features subsets and parameters of fuzzy SVM-based classifier. Finally, several simulation experiments and a real example are addressed to validate the feasibility and effectiveness of the proposed methodology. And the results of simulation experiments demonstrate that it can achieve excellent performance for CCPR and outperforms other approaches, such as learning vector quantization network, multi-layer perceptron network, probability neural network, fuzzy clustering and SVM, in term of recognition accuracy. The results of the practical cases manifest that the proposed method has application potential for solving the problem of control chart interpretation in real-world. [ABSTRACT FROM AUTHOR]

The number of traffic accidents in Brazil has reached alarming levels and is currently one of the leading causes of death in the country. With the number of vehicles on the roads increasing rapidly, these problems will tend to worsen. Consequently, huge investments in resources to increase road safety will be required. The vertical R-19 system for optical character recognition of regulatory traffic signs (maximum speed limits) according to Brazilian Standards developed in this work uses a camera positioned at the front of the vehicle, facing forward. This is so that images of traffic signs can be captured, enabling the use of image processing and analysis techniques for sign detection. This paper proposes the detection and recognition of speed limit signs based on a cascade of boosted classifiers working with haar-like features. The recognition of the sign detected is achieved based on the optimum-path forest classifier (OPF), support vector machines (SVM), multilayer perceptron, k-nearest neighbor (kNN), extreme learning machine, least mean squares, and least squares machine learning techniques. The SVM, OPF and kNN classifiers had average accuracies higher than 99.5 %; the OPF classifier with a linear kernel took an average time of 87 $$\upmu$$ s to recognize a sign, while kNN took 11,721 $$\upmu$$ s and SVM 12,595 $$\upmu$$ s. This sign detection approach found and recognized successfully 11,320 road signs from a set of 12,520 images, leading to an overall accuracy of 90.41 %. Analyzing the system globally recognition accuracy was 89.19 %, as 11,167 road signs from a database with 12,520 signs were correctly recognized. The processing speed of the embedded system varied between 20 and 30 frames per second. Therefore, based on these results, the proposed system can be considered a promising tool with high commercial potential. [ABSTRACT FROM AUTHOR]

Static security analysis is an important study carried out in the control centers of electric utilities. Static security assessment (SSA) is the process of determining whether the current operational state is in a secure or emergency (insecure) state. Conventional method of security evaluation involves performing continuous load flow analysis, which is highly time consuming and infeasible for real-time applications. This led to the application of pattern recognition (PR) approach for static security analysis. This paper presents a more efficient design of a PR system suitable for on-line SSA. The feature selection stage in the PR system uses many algorithms to select the optimal feature set. This paper proposes the use of Support Vector Machine (SVM), a recently introduced machine learning tool, in the classifier design stage of PR system. The developed PR system is implemented in IEEE standard test systems for SSA and classification. The performance of SVM classifier is compared with the conventional K-nearest neighbor, method of least squares and neural network classifiers. Simulation results prove that the SVM-PR classifier outperforms other equivalent classifier algorithms, giving high classification accuracy and less misclassification rate. The feasibility of SVM-PR classifier for on-line security assessment process is also presented. [ABSTRACT FROM AUTHOR]

This paper describes a trainable and flexible system able to recognize visual dynamic events, e.g. movements performed by different people, from a stream of images taken by a fixed camera. Each event is represented by a feature vector built from the spatio-temporal changes detected in the observed image sequence. The system neither attempts to recover the 3D structure nor assumes a prior model of the observed dynamic events. During training a supervisor identifies and labels the events of interest among those automatically detected by the system. At run time, previously unseen events are detected and classified on the basis of the available examples. Several experiments on real images are reported and the benefits of using Support Vector Machines for performing effective classification from a relatively small number of labeled examples and for building noise tolerant representations are discussed. Preliminary results indicate that the proposed system can also be applied with equally good results to the case in which the dynamic events are gestures performed by different people. [ABSTRACT FROM AUTHOR]

The World Health Organization suggests visual examination of stained sputum smear samples as a preliminary and basic diagnostic technique for diagnosing tuberculosis. The visual examination process requires much time of laboratorian, and also, it is prone to mistakes. For this purpose, this paper proposes a novel random forest (RF)-based segmentation and classification approaches for the automated classification of Mycobacterium tuberculosis in microscopic images of Ziehl-Neelsen-stained sputum smears obtained using a light-field microscope. The RF supervised learning method is improved to classify each pixel depending on local color distributions as a part of candidate bacilli regions. Therefore, each pixel is labeled as either a candidate tuberculosis (TB) bacilli pixel or not. The candidate pixels are grouped together using connected component analysis. Each pixel group is then rotated, resized and centrally positioned within a bounding box, respectively, in order to utilize appearance-based tuberculosis bacteria identification algorithms. Finally, each region is classified by using the proposed RF learning algorithm trained on manually marked TB bacteria regions in the training images. The algorithm produces results that agree well with manual segmentation and identification. Different two-class pixel and object classifiers are also compared to show the performance of the proposed RF-based pixel segmentation and bacilli objects identification algorithm. The sensitivity and specificity of the proposed classifier are above 75.77 and 96.97 % for the segmentation of the pixels, respectively. It is also revealed that the sensitivity increases over 93 % when the staining is performed in accordance with the procedure. Moreover, these measures are above 89.34 and 62.89 % for the identification of bacilli objects. The results show that the proposed novel method is quite successful when compared to the other applied methods. [ABSTRACT FROM AUTHOR]

In order to grasp and transport an object, grip and load forces must be scaled according to the object's properties (such as weight). To select the appropriate grip and load forces, the object weight is estimated based on experience or, in the case of robots, usually by use of image recognition. We propose a new approach that makes a robot's weight estimation less dependent on prior learning and, thereby, allows it to successfully grasp a wider variety of objects. This study evaluates whether it is feasible to predict an object's weight class in a replacement task based on the time series of upper body angles of the active arm or on object velocity profiles. Furthermore, we wanted to investigate how prediction accuracy is affected by (i) the length of the time series and (ii) different cross-validation (CV) procedures. To this end, we recorded and analyzed the movement kinematics of 12 participants during a replacement task. The participants' kinematics were recorded by an optical motion tracking system while transporting an object, 80 times in total from varying starting positions to a predefined end position on a table. The object's weight was modified (made lighter and heavier) without changing the object's visual appearance. Throughout the experiment, the object's weight (light/heavy) was randomly changed without the participant's knowledge. To predict the object's weight class, we used a discrete cosine transform to smooth and compress the time series and a support vector machine for supervised learning from the achieved discrete cosine transform parameters. Results showed good prediction accuracy (up to 95 % , depending on the CV procedure and the length of the time series). Even at the beginning of a movement (after only 300 ms), we were able to predict the object weight reliably (within a classification rate of 88 - 94 % ). [ABSTRACT FROM AUTHOR]

Motivated by the support vector data description, a classical one-class support vector machine, and the twin support vector machine classifier, this paper formulates a twin support vector hypersphere (TSVH) classifier, a novel binary support vector machine (SVM) classifier that determines a pair of hyperspheres by solving two related SVM-type quadratic programming problems, each of which is smaller than that of a conventional SVM, which means that this TSVH is more efficient than the classical SVM. In addition, the TSVH successfully avoids matrix inversion compared with the twin support vector machine, which indicates learning algorithms of the SVM can be easily extended to this TSVH. Computational results on several synthetic as well as benchmark data sets indicate that the proposed TSVH is not only faster, but also obtains better generalization. [ABSTRACT FROM AUTHOR]

This paper proposes a novel traffic event classification approach using event severities at intersections. The proposed system basically learns normal and common traffic flow by clustering vehicle trajectories. Common vehicle routes are generated by implementing trajectory clustering with Continuous Hidden Markov Model. Vehicle abnormality is detected by observing maximum likelihoods of partial vehicle locations and velocities on underlying common route models. The second part of the work is based on extracting the severities of abnormality by deviation measurement using Coefficient of Variances method. By using abnormal event samples, two severity classes are built in order to recognize event severities by Support Vector Machines and k-Nearest Neighborhood algorithms. Experimental results show that the proposed model has high precision with satisfactory incident detection and event severity classification performance. [ABSTRACT FROM AUTHOR]

Support vector machine (SVM) is one of the state-of-the-art tools for linear and nonlinear pattern classification. One of the design issues in SVM classifier is reducing the number of support vectors without compromising the classification accuracy. A technique denoted as diminishing learning (DL) is already proposed in literature for an SVM based multi-class isolated digit recognition system using speaker dependent TI46 database of isolated digits. In this paper, the computational complexity for SVM and SVM-DL based isolated digit recognition system is studied and the computation time for both the classifiers is evaluated by system-on-programmable-chip (SOPC) implementation of the recognition system onto an Altera Cyclone II Series FPGA using Nios II Soft-core processor. The number of support vectors is reduced by 38.28-90.25 % on using SVM-DL for isolated digit recognition problem. This in turn reduces the classification time for SVM-DL by 31.45-91.78 % over SVM. Recognition accuracies of 97 and 98 % are achieved for SVM classifier with and without DL technique, respectively. The study confirms the effect of, the order in which the classes are classified, on the recognition accuracy. For the TI46 database, about 2-4 % increase in recognition accuracy is obtained by choosing the optimum order for SVM-DL classifier. The proposed SOPC implementation of SVM-DL based recognition system can be employed for various other pattern recognition applications too such as face recognition, character recognition and target recognition. [ABSTRACT FROM AUTHOR]

The recently proposed twin support vector machine (TWSVM) obtains much faster training speed and comparable performance than classical support vector machine. However, it only considers the empirical risk minimization principle, which leads to poor generalization for real-world applications. In this paper, we formulate a robust minimum class variance twin support vector machine (RMCV-TWSVM). RMCV-TWSVM effectively overcomes the shortcoming in TWSVM by introducing a pair of uncertain class variance matrices in its objective functions. As a special case, we present a special type of the uncertain class variance matrices by combining the empirical positive and negative class variance matrices. Computational results on several synthetic as well as benchmark datasets indicate the significant advantages of proposed classifier in both computational time and test accuracy. [ABSTRACT FROM AUTHOR]

Graph-based methods have aroused wide interest in pattern recognition and machine learning, which capture the structural information in data into classifier design through defining a graph over the data and assuming label smoothness over the graph. Laplacian Support Vector Machine (LapSVM) is a representative of these methods and an extension of the traditional SVM by optimizing a new objective additionally appended Laplacian regularizer. The regularizer utilizes the local linear patches to approximate the data manifold structure and assumes the same label of the data on each patch. Though LapSVM has shown more effective classification performance than SVM experimentally, it in fact concerns more the locality than the globality of data manifold due to the Laplacian regularizer itself. As a result, LapSVM is relatively sensitive to the local change of the data and cannot characterize the manifold quite faithfully. In this paper, we design an alternative regularizer, termed as Glocalization Pursuit Regularizer. The new regularizer introduces a natural global structure measure to grasp the global and local manifold information as simultaneously as possible, which can be proved to make the representation of the manifold more compact than the Laplacian regularizer. We further introduce the new regularizer into SVM to develop an alternative graph-based SVM, called as Glocalization Pursuit Support Vector Machine (GPSVM). GPSVM not only inherits the advantages of both SVM and LapSVM but also uses the structural information more reasonably to guide the classifier design. The experiments both on the toy and real-world datasets demonstrate the better classification performance of our proposed GPSVM compared with SVM and LapSVM. [ABSTRACT FROM AUTHOR]

In the context of learning theory many efforts have been devoted to developing classification algorithms able to scale up with massive data problems. In this paper the complementary issue is addressed, aimed at deriving powerful classification rules by accurately learning from few data. This task is accomplished by solving a new mixed integer programming model that extends the notion of discrete support vector machines, in order to derive an optimal set of separating hyperplanes for binary classification problems. According to the cardinality of the set of hyperplanes, the classification region may take the form of a convex polyhedron or a polytope in the original space where the examples are defined. Computational tests on benchmark datasets highlight the effectiveness of the proposed model, that yields the greatest accuracy when compared to other classification approaches. [ABSTRACT FROM AUTHOR]

In toxicology, hazardous substances detected in organisms may often lead to different pathological conditions depending on the type of exposure and level of dosage; hence, further analysis on this can suggest the best cure. Urine profiling may serve the purpose because samples typically contain hundreds of compounds representing an effective metabolic fingerprint. This paper proposes a pattern recognition procedure for determining the type of cadmium dosage, acute or chronic, administrated to laboratory rats, where urinary profiles are detected using capillary electrophoresis. The procedure is based on the composition of a sample data matrix consisting of areas of common peaks, with appropriate pre-processing aimed at reducing the lack of reproducibility and enhancing the potential contribution of low-level metabolites in discrimination. The matrix is then used for pattern recognition including principal components analysis, cluster analysis, discriminant analysis and support vector machines. Attention is particularly focussed on the last of these techniques, because of its novelty and some attractive features such as its suitability to work with datasets that are small and/or have low samples/variable ratios. The type of cadmium administration is detected as a relevant feature that contributes to the structure of the sample matrix, and samples are classified according to the class membership, with discriminant analysis and support vector machines performing complementarily on a training and on a test set. [ABSTRACT FROM AUTHOR]

Anomaly detection in time-series data is a relevant problem in many fields such as stochastic data analysis, quality assurance, and predictive modeling. Markov models are an effective tool for time-series data analysis. Previous approaches utilizing Markov models incorporate transition matrices (TMs) at varying dimensionalities and resolutions. Other analysis methods treat TMs as vectors for comparison using search algorithms such as the nearest neighbors comparison algorithm, or use TMs to calculate the probability of discrete subsets of time-series data. We propose an analysis method that treats the elements of a TM as random variables, parameterizing them hierarchically. This approach creates a metric for determining the 'normalcy' of a TM generated from a subset of time-series data. The advantages of this novel approach are discussed in terms of computational efficiency, accuracy of anomaly detection, and robustness when analyzing sparse data. Unlike previous approaches, this algorithm is developed with the expectation of sparse TMs. Accounting for this sparseness significantly improves the detection accuracy of the proposed method. Detection rates in a variety of time-series data types range from (97 % TPR, 2.1 % FPR) to (100 % TPR, <0.1 % FPR) with very small sample sizes (20-40 samples) in data with sparse transition probability matrices. [ABSTRACT FROM AUTHOR]

The authors describe the application of two types of metallic foams modified with either graphene (GR) or carbon nanotubes (CNTs) as voltammetric electrodes in order to discriminate rice wines of different age and brand. Two types of bare metallic foams (bare Ni and Cu foam electrodes) were combined with GR or CNTs to give four types of modified metallic foams, referred to as GR/Ni, GR/Cu, CNT/Ni, and CNT/Cu foam electrodes. Cyclic voltammetry was applied to study the effects of GR and CNTs on the response of the electrodes. Multifrequency rectangle pulse voltammetry and multifrequency staircase pulse voltammetry were applied to generate potential waveforms, and chronoamperometric curves were recorded. Principal component analysis (PCA) allowed a classification of the rice wines, and characteristic regular distributions were identified in the PCA plots. Support vector machines (SVM) were found to perform better than partial least squares regression in predicting ages and brands of the rice wines in that all fit correlation coefficients were >0.9930. The SVM based leave-one-out cross-validation method proved to be the most powerful regression tool. The six types of foam electrodes perform very well in the classification and prediction of rice wines of different ages and brands. [ABSTRACT FROM AUTHOR]

To better identify the crack propagation pattern of coal or rock sample during loading, such as the initiation of new cracks or the propagation of pre-existing cracks. Acoustic emission (AE) characteristics of coal sample were analyzed to illustrate the correlation between AE frequency-domain information and crack propagation pattern. By introducing unsupervised and supervised algorithms to classify AE signals caused by crack pattern in different damage stages, and compare with the experimental damage. The results show that AE peak frequency, centroid frequency and weight frequency can be used to distinguish the initiation of new cracks or the propagation of pre-existing cracks. Fuzzy C-Means (FCM) unsupervised method was used to cluster AE signals generated by specific crack propagation pattern, and the maximum error was 3.8%. In addition, we compared the performance of Support Vector Machine (SVM), Extreme Learning Machine (ELM) and Random Forest (RF) algorithms, the accuracy of RF was higher than ELM, and the running time was significantly 20 times lower than SVM. After expanding the application of RF, it was consistent with the experimental observation. This study contributes to evaluate the crack initiation, propagation and location of engineering structures, such as mine coal pillar and bridge. Highlights: The correlation between acoustic emission waveform features and crack modes is analyzed. A framework is designed to obtain coal fracture types from AE signals based on multiple machine learning methods. Random forest is the most suitable method, which is in good agreement with the experimental results. [ABSTRACT FROM AUTHOR]

Extraction of built-up area (BUA) is essential for proper city planning and management. It enables the concerned authorities to formulate better city development policies and manage emergent disasters. However, the traditionally used optical data present spectral confusion where BUAs are mixed with other features adding to management complexities. Therefore, an advanced automated method is required to extract the spectral and textural features from satellite data for the pattern recognition of BUA. Landsat-8 Operational Land Imager (OLI) has been used in the current study to identify the pattern and extract BUA of Gujranwala, Pakistan. First, eight textural features based on the gray-level co-occurrence matrix (GLCM) are selected and combined with multispectral data. Then, feature selection methods are applied to select optimal features used to train the proposed support vector machine (SVM) classifier. Finally, the results from SVM classifiers are compared with k-nearest neighbor (k-NN) and backpropagation neural network (BP-NN) to highlight any improvements in results. The comparisons show that the proposed approach increases the overall accuracy of linear-SVM by 8.41%, radial basis function SVM by 8.3%, BP-NN by 7.63%, and k-NN by 6.6%. This can help city managers and planners to extract critical BUA information in otherwise unplanned and rapidly expanding cities to move toward smart and sustainable cities. [ABSTRACT FROM AUTHOR]

Multi-attribute motion data can be generated in many applications/ devices, such as motion capture devices and animations. It can have dozens of attributes, thousands of rows, and even similar motions can have different durations and different speeds at corresponding parts. There are no row-to-row correspondences between data matrices of two motions. To be classified and recognized, multi-attribute motion data of different lengths are reduced to feature vectors by using the properties of singular value decomposition (SVD) of motion data. The reduced feature vectors of similar motions are close to each other, while reduced feature vectors are different from each other if their motions are different. By applying support vector machines (SVM) to the feature vectors, we efficiently classify and recognize real-world multi-attribute motion data. With our data set of more than 300 motions with different lengths and variations, SVM outperforms classification by related similarity measures, in terms of accuracy and CPU time. The performance of our approach shows its feasibility of real-time applications to real-world data. [ABSTRACT FROM AUTHOR]

Characteristics like robustness, adaptability to several load conditions, and low costs of operation and maintenance are some of the reasons three-phase induction motors are ubiquitous in industrial applications. Even so, these machines are subjected to electrical and mechanical faults which can result in malfunction. The detection of incipient faults is the focus of several recent studies because they provide information for timely decisions to avoid unplanned stops of industrial processes. A common problem of induction motors is the presence of broken rotor bars. Most of the methods of condition monitoring and fault diagnosis focus on the usage of only one type of power supply: line-connected or fed by frequency inverters. Given this situation, we present an alternative technique for the detection of broken rotor bars of three-phase induction motors regardless of the type of power supply. Our approach is based on similarity measures of the stator current signals of two motor phases in order to extract the relevant features of such signals, which are then classified using three intelligent systems: artificial neural network, support vector machines, and k-nearest neighbors. We performed 3310 experimental tests with motors operating in steady state, with sinusoidal and non-sinusoidal power supply, and variations of voltage unbalance levels, supply frequency, and load torque. Classification accuracy rates over 92% of success were obtained in these tests, which validate the proposed approach. [ABSTRACT FROM AUTHOR]

Imbalanced classification has been a major challenge for machine learning because many standard classifiers mainly focus on balanced datasets and tend to have biased results toward the majority class. We modify entropy fuzzy support vector machine (EFSVM) and introduce instance-based entropy fuzzy support vector machine (IEFSVM). Both EFSVM and IEFSVM use the entropy information of k-nearest neighbors to determine the fuzzy membership value for each sample which prioritizes the importance of each sample. IEFSVM considers the diversity of entropy patterns for each sample when increasing the size of neighbors, k, while EFSVM uses single entropy information of the fixed size of neighbors for all samples. By varying k, we can reflect the component change of sample's neighbors from near to far distance in the determination of fuzzy value membership. Numerical experiments on 35 public and 12 real-world imbalanced datasets are performed to validate IEFSVM, and area under the receiver operating characteristic curve (AUC) is used to compare its performance with other SVMs and machine learning methods. IEFSVM shows a much higher AUC value for datasets with high imbalance ratio, implying that IEFSVM is effective in dealing with the class imbalance problem. [ABSTRACT FROM AUTHOR]

Twin support vector machine (TWSVM) is proved to be better than support vector machine (SVM) in most cases, since it only deals with two smaller quadratic programming problems, which leads to high computational efficiency. It is proposed to solve a single-task learning problem, just like many other machine learning algorithms. However, a learning task may have relationships with other tasks in many practical problems. Training those tasks independently may neglect the underlying information among all tasks, while such information may be useful to improve the overall performance. Inspired by the multi-task learning theory, we propose two novel multi-task ν -TWSVMs. Both models inherit the merits of multi-task learning and ν -TWSVM. Meanwhile, they overcome the shortcomings of other multi-task SVMs and multi-task TWSVMs. Experimental results on three benchmark datasets and two popular image datasets also clearly demonstrate the effectiveness of our methods. [ABSTRACT FROM AUTHOR]

Partial discharge (PD) measurement is used for a diagnosis and performance assessment of the solid insulation material inside the high-voltage (HV) equipment. PD measurement indicates the presence of voids, cracks and imperfection present in solid insulation material. The major problem associated with this measured PD signal is heavily contaminated by noise which results in reduction in PD pattern recognition. The objective of this work is to measure and de-noise the PD signal due to cavity and recognize two different size of cavities present in three different locations, namely near HV electrode, center and lower electrode. In first part, the measured PD signal is de-noised using translation invariant wavelet transform. In second part, the three-dimensional (φ–q–n) PD patterns are extracted from the de-noised PD data. Then, it is subjected to canny edge detection technique, and the features like horizontal and vertical fractal dimension averages are evaluated using fractal image compression-based semi-variance technique. For classification, multi-class nonlinear support vector machine has been proposed to classify position and size of the cavity based on the PD fingerprints. The findings of this proposed work can be used to design a solid basis for an recognition of cavity size and position in an electrical apparatus. [ABSTRACT FROM AUTHOR]

The application of Poincaré plot analysis to characterize inter-beat interval dynamics has been successfully proposed in the scientific literature for the assessment of humans' physiological states and related aberrations. In this study, we proposed novel descriptors to trace the evolution of Poincaré plot shape over the lags. Their reliability in ultra-short cardiovascular series analysis was validated on synthetic inter-beat series generated through a physiologically plausible integral pulse frequency modulation model. Furthermore, we used the proposed approach for the investigation of the direct relationship between autonomic nervous system (ANS) dynamics and hedonic olfactory elicitation, in a group of 30 healthy subjects. Participants with a similar olfactory threshold were selected, and were asked to score 5-s stimuli in terms of arousal and valence levels according to the Russell's circumflex model of affect. Their ANS response was investigated in 35-s windows after the elicitation. Experimental results showed a gender-specific, high discriminant power of the proposed approach, discerning between pleasant and unpleasant odorants with an accuracy of 83.33% and 73.33% for men and for women, respectively. Graphical Abstract Olfaction plays a crucial role in our life and is strictly related to the Autonomic Nervous System (ANS) activity, which can be monitored studying Heart Rate Variability. We used the Lagged Poincare Plot approach to recognize gender-specific ANS response in 35-second windows after the elicitation through pleasant/unpleasant odorants. [ABSTRACT FROM AUTHOR]

Unconventional oil and gas reservoirs from the lower Palaeozoic basin at the western slope of the East European Craton were taken into account in this study. The aim was to supply and improve standard well logs interpretation based on machine learning methods, especially ANNs. ANNs were used on standard well logging data, e.g. P-wave velocity, density, resistivity, neutron porosity, radioactivity and photoelectric factor. During the calculations, information about lithology or stratigraphy was not taken into account. We apply different methods of classification: cluster analysis, support vector machine and artificial neural network—Kohonen algorithm. We compare the results and analyse obtained electrofacies. Machine learning method–support vector machine SVM was used for classification. For the same data set, SVM algorithm application results were compared to the results of the Kohonen algorithm. The results were very similar. We obtained very good agreement of results. Kohonen algorithm (ANN) was used for pattern recognition and identification of electrofacies. Kohonen algorithm was also used for geological interpretation of well logs data. As a result of Kohonen algorithm application, groups corresponding to the gas-bearing intervals were found. Analysis showed diversification between gas-bearing formations and surrounding beds. It is also shown that internal diversification in gas-saturated beds is present. It is concluded that ANN appeared to be a useful and quick tool for preliminary classification of members and gas-saturated identification. [ABSTRACT FROM AUTHOR]

This article presents a computer vision system for real-time detection and robust recognition of speed limit signs, specially designed for intelligent vehicles. First, a new segmentation method is proposed to segment the image, and the CHT transformation (circle hog transform) is used to detect circles. Then, a new method based on local binary patterns is proposed to filter segmented images in order to reduce false alarms. In the classification phase, a cascading architecture of two linear support vector machines is proposed. The first is trained on the GTSRB dataset to decide whether the detected region is a speed limit sign or not, and the second is trained on the MNIST dataset to recognize the sign numbers. The system achieves a classification recall of 99.81% with a precision of 99.08% on the GTSRB dataset; in addition, the system is also tested on the BTSD and STS datasets, and it achieves a classification recall of 99.39% and 98.82% with a precision of 99.05% and 98.78%, respectively, within a processing time of 11.22 ms. [ABSTRACT FROM AUTHOR]

Breast cancer is one of the leading causes of death by cancer among women. The high mortality rates and the occurrence of this cancer worldwide show the importance of the investigation and development of means for the detection and early diagnosis of this disease. Computer-Aided Detection and Diagnosis systems have been developed to improve diagnostic accuracy by radiologists. This work proposes a method for discriminating patterns of malignancy and benignity of masses in digitized mammography images through the analysis of local features. The method comparatively applies the Scale-Invariant Feature Transform (SIFT), Speed Up Robust Feature (SURF), Oriented Fast and Rotated BRIEF (ORB) and Local Binary Pattern (LBP) descriptors for local feature extraction. These features are represented by a Bag of Features (BoF) model, applied to provide new representations of the data and to reduce its dimensionality. Finally, the features are used as input for the Support Vector Machine (SVM), Adaptive Boosting (Adaboost) and Random Forests (RF) classifiers to differentiate malignant and benign masses. The method obtained significant results, reaching 100% sensitivity, 99.65% accuracy and 99.24% specificity for benign and malignant mass classification. [ABSTRACT FROM AUTHOR]

This work presents automatic identification and verification approaches based on lip biometrics, using a static lip shape and applying a lip correction preprocessing, transforming data from Hidden Markov Model and being classified by Support Vector Machines. The classification system is conclusive for the identification of a person by the shape of the lips, even if the person presents soft facial emotions. Moreover the use of static lips shape has been revealed as a good option for security applications. The experiments have been carried out with two public datasets. One dataset was used to model and validate the approach, and the other dataset has been used to test the model blindly. The accuracy is up to 100% and 99.76% for GDPS-ULPGC and RaFD datasets respectively, using two training samples under a hold-out validation. Based on the results we can conclude that the system is very robust and stable with the highest classification capacity and minimal computation complexity. [ABSTRACT FROM AUTHOR]

Arabic Text document clustering is an important aspect for providing conjectural navigation and browsing techniques by organizing massive amounts of data into a small number of defined clusters. However, Words in form of vector are used for clustering methods is often unsatisfactory as it ignores relationships between important terms. Cluster analysis separates data into groups on clusters for the purposes of improved understanding or summarization. Clustering has a long history and many techniques developed in statistics, data mining, pattern recognition and other fields. This research proposes three approaches; Unsupervised, Semi Supervised techniques and Semi Supervised with dimensionality reduction to construct a clustering based classifier for Arabic text documents. Using k-means, incremental k-means, Threshold + k-means and k-means with dimensionality reduction, after document preprocessing removing stop words and gets the root for each term in each document. Then apply a term weighting method to get the weight of each term with respect to its document. Then apply a similarity measure method to each document and its similarity with other documents. And using F-measure, entropy and support vector machine (SVM) for calculate accuracy. The datasets are online dynamic datasets that are characterized by its availability and credibility on the internet. Arabic language is a challenging language when applied in an inference based algorithm. So, selecting the appropriate dataset is a principal factor in such research. The accuracy of those methods compared with other approaches and the proposed methods shows better accuracy and fewer errors for new classification test cases. Considering that the dimension reduction process is very sensitive because increasing the ratio of reduction can destroy important terms. [ABSTRACT FROM AUTHOR]

The article contains a description of two quantum circuits for pattern recognition. The first approach is realized with use of k nearest neighbors algorithm and the second with support vector machine. The task is to distinguish between Hermitian and non-Hermitian matrices. The quantum circuits are constructed to accumulate elements of a learning set. After this process, circuits are able to produce a quantum state which contains the information if a tested element fits to the trained pattern. To improve the efficiency of presented solutions, the matrices were uniquely labeled with feature vectors. The role of the feature vectors is to highlight some features of the objects which are crucial in the process of classification. The circuits were implemented in Python programming language and some numeric experiments were conducted to examine the capacity of presented solutions in pattern recognition. [ABSTRACT FROM AUTHOR]

Multiple instance learning attempts to learn from a training set consists of labeled bags each containing many unlabeled instances. In previous works, most existing algorithms mainly pay attention to the 'most positive' instance in each positive bag, but ignore the other instances. For utilizing these unlabeled instances in positive bags, we present a new multiple instance learning algorithm via semi-supervised laplacian twin support vector machines (called Miss-LTSVM). In Miss-LTSVM, all instances in positive bags are used in the manifold regularization terms for improving the performance of classifier. For verifying the effectiveness of the presented method, a series of comparative experiments are performed on seven multiple instance data sets. Experimental results show that the proposed method has better classification accuracy than other methods in most cases. [ABSTRACT FROM AUTHOR]

Twin support vector machine (TWSVM) is regarded as a milestone in the development of powerful SVMs. However, there are some inconsistencies with TWSVM that can lead to many reasonable modifications with different outputs. In order to obtain better performance, we propose a novel combined outputs framework that combines rational outputs. Based on this framework, an optimal output model, called the linearly combined twin bounded support vector machine (LCTBSVM), is presented. Our LCTBSVM is based on the outputs of several TWSVMs, and produces the optimal output by solving an optimization problem. Furthermore, two heuristic algorithms are suggested in order to solve the optimization problem. Our comprehensive experiments show the superior generalization performance of our LCTBSVM compared with SVM, PSVM, GEPSVM, and some current TWSVMs, thus confirming the value of our theoretical analysis approach. [ABSTRACT FROM AUTHOR]

Strabismus is a pathology that affects approximately 4 % of the population, causing aesthetic problems reversible at any age and irreversible sensory alterations that modify the vision mechanism. The Hirschberg test is one type of examination for detecting this pathology. Computer-aided detection/diagnosis is being used with relative success to aid health professionals. Nevertheless, the routine use of high-tech devices for aiding ophthalmological diagnosis and therapy is not a reality within the subspecialty of strabismus. Thus, this work presents a methodology to aid in diagnosis of syndromic strabismus through digital imaging. Two hundred images belonging to 40 patients previously diagnosed by an specialist were tested. The method was demonstrated to be 88 % accurate in esotropias identification (ET), 100 % for exotropias (XT), 80.33 % for hypertropias (HT), and 83.33 % for hypotropias (HoT). The overall average error was 5.6 Δ and 3.83 Δ for horizontal and vertical deviations, respectively, against the measures presented by the specialist. [ABSTRACT FROM AUTHOR]