Wyniki wyszukiwania - BazTech

1

Memory response on generalized thermoelastic medium in context of dual phase lag thermoelasticity with non-local effect

Tiwari R., Saeed A. M., Kumar R., Kumar A., Singhal A.

Archives of Mechanics

|

2022

|

Vol. 74, nr 2-3

69--88

EN

Theory of non-local continuum is contemporary appraised and is found to be supplementary coherent to capture the impacts of each and every point of the material at its single point. The conviction of memory dependent derivative is also newly appraised and is observed to be more intuitionistic for predicting the realistic character of the real-world obstacles. Attractiveness of the belief of a memory dependent derivative lies in its unique properties such as its significant constituents – a kernel function and time-delay are freely selected according to the requirement of a problem. The present study comprises a new meticulous thermoelastic heat conduction model for the homogeneous, isotropic, thermoelastic half space medium concerning memory effects and non-local effects. Governing equations are constructed on the basis of the newly appraised non-local generalized theory of thermoelasticity with two phase lags in the frame of a memory dependent derivative. Exact analytical solutions of the physical fields such as dimensionless temperature, displacement as well as thermal stress are evaluated by using a suitable technique of the Laplace transform. Quantitative results are determined in a time-domain for different values of time by taking the numerical inversion of the Laplace transform. Noteworthy role of the constituents of the memory dependent derivative such as kernel function as well as time-delay factor has been scrutinized on the crucial field variables of the medium through computational outcomes. Moreover, the impact of non-local parameter is examined on the variations of field quantities through the quantitative results.

2

Simulating Caspian Sea surface water level by artifcial neural network and support vector machine models

Khaledian M. R., Isazadeh M., Biazar S. M., Pham Q. B.

Acta Geophysica

|

2020

|

Vol. 68, no. 2

553--563

EN

Reduction in sea water level can make services in nearshore structures difcult, and sea water level rise increases the risk to residential areas or the surrounding felds. For strategic planning, it is vital to take into account the present and future fuctuations of Caspian Sea water level. In this study, support vector machine and artifcial neural network are used to estimate water level of the Caspian Sea. A 34-year period dataset is used as input data for water level on the scale based at Anzali, Iran. Performances of these two models are compared according to some statistical indices. Results of this study indicate that support vector machine with an error of 4.782 mm and r=0.96 simulated the time series better, as compared with artifcial neural network with an error of 5.014 mm and r=0.957; furthermore, the uncertainty of this model is lower than that of the artifcial neural network, i.e., 0.04 verses 0.22.

3

On the Existence of Kernel Function for Kernel-Trick of k-Means in the Light of Gower Theorem

Kłopotek Mieczysław A.

Fundamenta Informaticae

|

2019

|

Vol. 168, nr 1

25--43

EN

This paper, constituting an extension to the conference paper [1], corrects the proof of the Theorem 2 from the Gower’s paper [2, page 5]. The correction is needed in order to establish the existence of the kernel function used commonly in the kernel trick e.g. for k-means clustering algorithm, on the grounds of distance matrix. The correction encompasses the missing if-part proof and dropping unnecessary conditions.

4

A hybrid approach for the delineation of brain lesion from CT images

Gautam A., Raman B., Raghuvanshi S.

Biocybernetics and Biomedical Engineering

|

2018

|

Vol. 38, no. 3

504--518

EN

Brain lesion segmentation from radiological images is the most important task in accurate diagnosis of patients. This paper presents a hybrid approach for the segmentation of brain lesion from computed tomography (CT) images based on the combination of fuzzy clustering using hyper tangent function as the robust kernel and distance regularized level set evolution (DRLSE) function as the edge based active contour method. Kernel based fuzzy clustering method divides the image into different regions. These regions can be used to find region of interest by using DRLSE algorithm to generate the optimal region boundary. The proposed method results in smooth boundary of the required regions with high accuracy of segmentation. In this paper, results are compared with standard fuzzy c-means (FCM) clustering, spatial FCM, robust kernel based fuzzy clustering (RFCM) and DRLSE algorithms. The performance of the proposed method is evaluated on CT scan images of hemorrhagic lesion, which shows that our method can segment brain lesion more accurately than the other conventional methods.

5

Auto-kernel using multilayer perceptron

Cheng W. C.

Journal of Theoretical and Applied Computer Science

|

2012

|

Vol. 6, nr 2

60-71

EN

This work presents a constructive method to train the multilayer perceptron layer after layer successively and to accomplish the kernel used in the support vector machine. Data in different classes will be trained to map to distant points in each layer. This will ease the mapping of the next layer. A perfect mapping kernel can be accomplished successively. Those distant mapped points can be discriminated easily by a single perceptron.

6

A Robust Multiple Classifier System for Pixel Classification of Remote Sensing Images

Maulik U., Chakraborty D.

Fundamenta Informaticae

|

2010

|

Vol. 101, nr 4

286-304

EN

Satellite image classification is a complex process that may be affected by many factors. This article addresses the problem of pixel classification of satellite images by a robust multiple classifier system that combines k-NN, support vector machine (SVM) and incremental learning algorithm (IL). The effectiveness of this combination is investigated for satellite imagery which usually have overlapping class boundaries. These classifiers are initially designed using a small set of labeled points. Combination of these algorithms has been done based on majority voting rule. The effectiveness of the proposed technique is first demonstrated for a numeric remote sensing data described in terms of feature vectors and then identifying different land cover regions in remote sensing imagery. Experimental results on numeric data as well as two remote sensing data show that employing combination of classifiers can effectively increase the accuracy label. Comparison is made with each of these single classifiers in terms of kappa value, accuracy, cluster quality indices and visual quality of the classified images.

7

Performance of the Support Vector Machines for medical classification problems

Ćwiklińska-Jurkowska M.

Biocybernetics and Biomedical Engineering

|

2009

|

Vol. 29, no. 4

63-81

EN

In the Support Vector Machines classification technique the best possible discriminating hyperplane between two populations is looked for by maximizing of margin between the populations' closest points. This idea is also applied for obtaining nonlinear discriminant boundaries by using different kernels for transformations, thus obtaining a nonlinear Support Vector Machines method. The nonlinear Support Vector Machines method is based on pre-processing of data to represent patterns in high dimension- usually much higher than the original variable feature space. In the presented work the dependency of Support Vector Machines performance on the kind of kernel and Support Vector Machines parameters is presented. The performance was assessed by resubstitution, 10- fold cross-validation, leave-one-out error, learning curves and Receiver Operating Characteristic curves. The kind and shape of the kernel is more important than regularization constant allowing different levels of overlapping classes. Combining boosting and Support Vector Machines did not improved performance in comparison to Support Vector Machines method alone, because both Support Vector Machines procedure and boosting are focused on observations difficult to classify.