scholarly journals A Linear-RBF Multikernel SVM to Classify Big Text Corpora

2015 ◽  
Vol 2015 ◽  
pp. 1-14 ◽  
Author(s):  
R. Romero ◽  
E. L. Iglesias ◽  
L. Borrajo

Support vector machine (SVM) is a powerful technique for classification. However, SVM is not suitable for classification of large datasets or text corpora, because the training complexity of SVMs is highly dependent on the input size. Recent developments in the literature on the SVM and other kernel methods emphasize the need to consider multiple kernels or parameterizations of kernels because they provide greater flexibility. This paper shows a multikernel SVM to manage highly dimensional data, providing an automatic parameterization with low computational cost and improving results against SVMs parameterized under a brute-force search. The model consists in spreading the dataset into cohesive term slices (clusters) to construct a defined structure (multikernel). The new approach is tested on different text corpora. Experimental results show that the new classifier has good accuracy compared with the classic SVM, while the training is significantly faster than several other SVM classifiers.

Author(s):  
Eduardo de la Guerra Ochoa ◽  
Javier Echávarri Otero ◽  
Enrique Chacón Tanarro ◽  
Benito del Río López

This article presents a thermal resistances-based approach for solving the thermal-elastohydrodynamic lubrication problem in point contact, taking the lubricant rheology into account. The friction coefficient in the contact is estimated, along with the distribution of both film thickness and temperature. A commercial tribometer is used in order to measure the friction coefficient at a ball-on-disc point contact lubricated with a polyalphaolefin base. These data and other experimental results available in the bibliography are compared to those obtained by using the proposed methodology, and thermal effects are analysed. The new approach shows good accuracy for predicting the friction coefficient and requires less computational cost than full thermal-elastohydrodynamic simulations.


2020 ◽  
Vol 190 (3) ◽  
pp. 342-351
Author(s):  
Munir S Pathan ◽  
S M Pradhan ◽  
T Palani Selvam

Abstract In the present study, machine learning (ML) methods for the identification of abnormal glow curves (GC) of CaSO4:Dy-based thermoluminescence dosimeters in individual monitoring are presented. The classifier algorithms, random forest (RF), artificial neural network (ANN) and support vector machine (SVM) are employed for identifying not only the abnormal glow curve but also the type of abnormality. For the first time, the simplest and computationally efficient algorithm based on RF is presented for GC classifications. About 4000 GCs are used for the training and validation of ML algorithms. The performance of all algorithms is compared by using various parameters. Results show a fairly good accuracy of 99.05% for the classification of GCs by RF algorithm. Whereas 96.7% and 96.1% accuracy is achieved using ANN and SVM, respectively. The RF-based classifier is recommended for GC classification as well as in assisting the fault determination of the TLD reader system.


2020 ◽  
Vol 10 (1) ◽  
Author(s):  
Alessio Lugnan ◽  
Emmanuel Gooskens ◽  
Jeremy Vatin ◽  
Joni Dambre ◽  
Peter Bienstman

AbstractMachine learning offers promising solutions for high-throughput single-particle analysis in label-free imaging microflow cytomtery. However, the throughput of online operations such as cell sorting is often limited by the large computational cost of the image analysis while offline operations may require the storage of an exceedingly large amount of data. Moreover, the training of machine learning systems can be easily biased by slight drifts of the measurement conditions, giving rise to a significant but difficult to detect degradation of the learned operations. We propose a simple and versatile machine learning approach to perform microparticle classification at an extremely low computational cost, showing good generalization over large variations in particle position. We present proof-of-principle classification of interference patterns projected by flowing transparent PMMA microbeads with diameters of $${15.2}\,\upmu \text {m}$$ 15.2 μ m and $${18.6}\,\upmu \text {m}$$ 18.6 μ m . To this end, a simple, cheap and compact label-free microflow cytometer is employed. We also discuss in detail the detection and prevention of machine learning bias in training and testing due to slight drifts of the measurement conditions. Moreover, we investigate the implications of modifying the projected particle pattern by means of a diffraction grating, in the context of optical extreme learning machine implementations.


Author(s):  
Hidetoshi Tanaka ◽  

Multiclassification problems are often binarized into pairwise classifications to use basic classification such as support vector machines (SVM). Instead of the widely used aggregation by fuzzy logical product, we propose simple double-negative aggregation, in which the membership functions use margin areas of SVM discrimination functions, and memberships of negative votes of the class are accumulated to produce the negative membership of the class. This provides results consistent with basic pairwise memberships, enumerates candidates when the total membership of multiple classes is nearly equal, and requires low computational cost in class reconfiguration.


Author(s):  
Mohand Lokman Ahmad Al-dabag ◽  
Haider Th. Salim ALRikabi ◽  
Raid Rafi Omar Al-Nima

One of the common types of arrhythmia is Atrial Fibrillation (AF), it may cause death to patients. Correct diagnosing of heart problem through examining the Electrocardiogram (ECG) signal will lead to prescribe the right treatment for a patient. This study proposes a system that distinguishes between the normal and AF ECG signals. First, this work provides a novel algorithm for segmenting the ECG signal for extracting a single heartbeat. The algorithm utilizes low computational cost techniques to segment the ECG signal. Then, useful pre-processing and feature extraction methods are suggested. Two classifiers, Support Vector Machine (SVM) and Multilayer Perceptron (MLP), are separately used to evaluate the two proposed algorithms. The performance of the last proposed method with the two classifiers (SVM and MLP) show an improvement of about (19% and 17%, respectively) after using the proposed segmentation method so it became 96.2% and 97.5%, respectively.


2016 ◽  
Vol 21 (3) ◽  
pp. 69-79 ◽  
Author(s):  
Abdelkhalek Bakkari ◽  
Anna Fabijańska

Abstract In this paper, the problem of segmentation of 3D Magnetic Resonance Imaging (MRI) and Computed Tomography (CT) brain images is considered. A supervoxel-based segmentation is regarded. In particular, a new approach called Relative Linear Interactive Clustering (RLIC) is introduced. The method, dedicated to image division into super-voxels, is an extension of the Simple Linear Interactive Clustering (SLIC) super-pixels algorithm. During RLIC execution firstly, the cluster centres and the regular grid size are initialized. These are next clustered by Fuzzy C-Means algorithm. Then, the extraction of the super-voxels statistical features is performed. The method contributes with 3D images and serves fully volumetric image segmentation. Five cases are tested demonstrating that our Relative Linear Interactive Clustering (RLIC) is apt to handle huge size of images with a significant accuracy and a low computational cost. The results of applying the suggested method to segmentation of the brain tumour are exposed and discussed.


Historical documents are important source for knowing culture, language, social activities, educational system, etc. The historical documents are in different languages and evolved over centuries and transformed to present modern language, classification of documents into various eras, recognition of words etc. In this paper, we have proposed a new approach to automatic identification of the age of the historical handwritten document images based on LBP (Local Binary Pattern) and LPQ (Local Phase Quantization) algorithm. The standard historical handwritten document images named as MPS (Medieval Paleographic Scale) dataset which is publicly available is used to experiment. LBP and LPQ descriptors are used to extract the features of the historical document images. Further, documents are classified based on the discriminating feature values using classifiers namely K-NN (K-Nearest Neighbors) and SVM (Support Vector Machine) classifier. The accuracy of historical handwritten document images by K-NN and SVM are 90.7% and 92.8% respectively.


2019 ◽  
Vol 12 (2) ◽  
pp. 549-562
Author(s):  
Alpika Tripathi ◽  
Geetika Srivastava ◽  
K.K. Singh ◽  
P.K. Maurya

The objective of this paper is to make a distinction between EEG data of normal and epileptic subjects. Methods: The dataset is taken from 20-30 years healthy male/female subjects from EEG lab of Dept. of Neurology, Dr. RML Institute of Medical Sciences, Lucknow (India). The feature extraction has been done using the Hilbert Huang Transform (HHT) method. The experimental EEG signals have been decomposed till 5th level of Intrinsic Mode Function (IMF) followed by calculation of high order statistical values of each IMF. Relief algorithm (RBAs) is used for feature selection and classification is performed using Linear Support Vector Machine (Linear SVM). This paper gives an independent approach of classifying Epileptic EEG data with reduced computational cost and high accuracy. Our classification result shows sensitivity, specificity, selectiv­ity and accuracy of 96.4%, 79.16%, 84.3% and 88.5% respectively. The proposed method has been analyzed to be very effective in accurate classification of epileptic EEG data with high sensitivity.


2020 ◽  
Vol 10 (14) ◽  
pp. 4991
Author(s):  
Carlos Villaseñor ◽  
Alberto A. Gallegos ◽  
Javier Gomez-Avila ◽  
Gehová López-González ◽  
Jorge D. Rios ◽  
...  

Environment classification is one of the most critical tasks for Unmanned Aerial Vehicles (UAV). Since water accumulation may destabilize UAV, clouds must be detected and avoided. In a previous work presented by the authors, Superpixel Segmentation (SPS) descriptors with low computational cost are used to classify ground, sky, and clouds. In this paper, an enhanced approach to classify the environment in those three classes is presented. The proposed scheme consists of a Convolutional Neural Network (CNN) trained with a dataset generated by both, an human expert and a Support Vector Machine (SVM) to capture context and precise localization. The advantage of using this approach is that the CNN classifies each pixel, instead of a cluster like in SPS, which improves the resolution of the classification, also, is less tedious for the human expert to generate a few training samples instead of the normal amount that it is required. This proposal is implemented for images obtained from video and photographic cameras mounted on a UAV facing in the same direction of the vehicle flight. Experimental results and comparison with other approaches are shown to demonstrate the effectiveness of the algorithm.


Sign in / Sign up

Export Citation Format

Share Document