Kernel Parameter Values Research Articles

Kernel Error Path Algorithm.

Tuning the values of kernel parameters plays a vital role in the performance of kernel methods. Kernel path algorithms have been proposed for several important learning algorithms, including support vector machine and kernelized Lasso, which can fit the piecewise nonlinear solutions of kernel methods with respect to the kernel parameter in a continuous space. Although the error path algorithms have been proposed to ensure that the model with the minimum cross validation (CV) error can be found, which is usually the ultimate goal of model selection, they are limited to piecewise linear solution paths. To address this problem, in this article, we extend the classic error path algorithm to the nonlinear kernel solution paths and propose a new kernel error path algorithm (KEP) that can find the global optimal kernel parameter with the minimum CV error. Specifically, we first prove that error functions of binary classification and regression problems are piecewise constant or smooth w.r.t. the kernel parameter. Then, we propose KEP for support vector machine and kernelized Lasso and prove that it guarantees to find the model with the minimum CV error within the whole range of kernel parameter values. Experimental results on various datasets show that our KEP can find the model with minimum CV error with less time consumption. Finally, it would have better generalization error on the test set, compared with grid search and random search.

Design and Experimental Research of Intelligent Inspection and Classification System for Yuba Skin Quality

At present, the surface quality of Yuba skin is determined by sensory methods. In order to realize the intelligent classification detection of Yuba skin quality, this study designed a system that automatically determines the quality of Yuba skin surfaces based on image processing and support vector machine (SVM) approaches. Specifically, the system uses image preprocessing to extract the grayscale eigenvalues, gray level co-occurrence matrix (GLCM) eigenvalues, and gray level run length matrix (GLRLM) eigenvalues of the sample image and uses them as input values for a quality grading system. Through model evaluation of three classification models, the SVM classification model was selected according to the evaluation results, and different kernel functions were used in the model for sample training. Based on Matlab, the quality grading software of Yuba skin was developed and designed. Intelligent detection and grading were realized through the radial basis kernel function support vector machine (RBF-SVM) grading model. The best penalty factor (c = 3.50) and kernel parameter value (g = 0.98) were obtained through cross-validation. The accuracy of the model was 95.31% and 94.16% for the training and test sets, respectively. The grading accuracy of the RBF-SVM grading system was 93.56%, and the error was less than 5% compared with the traditional sensory method of grading; thus, the quality classification method based on the SVM classification system for Yuba skin is feasible and can be used for quality detection.

Hybrid modified weighted water cycle algorithm and Deep Analytic Network for forecasting and trend detection of forex market indices

This paper presents forecasting and trend analysis of foreign currency exchange rate in financial market using a hybrid Deep Analytic Network (DAN) technique optimized by a modified water cycle algorithm called Weighted WCA (WWCA) with better generalization capability than the traditional WCA.DAN comprises several stacked KRR (Kernel Ridge Regression) Auto encoders in a multilayer nonlinear regression architecture approach that provides better generalization and accuracy using regularized least squares technique. Further DAN using wavelet kernel function is particularly attractive for its strong data fitting and generalization ability along with its simplified execution procedure, high speed, and better performance achievements in comparison to LSSVM (least squares support vector machine). The output from the DAN is fed to a weighted KRR module to reject noise or the outliers in the noisy data and to make DAN a more robust predictor of the Forex markets, To obtain optimal values of wavelet kernel parameters, a modified metaheuristic water cycle algorithm i.e. the proposed WWCA is utilized. Applications of this new approach to predict forex rate along with trend analysis on three stock markets provide successful results and validate its superiority over some well known approaches like ANN, SVM, Naïve-Bayes, ELM.

Adaptive Max-Margin One-Class Classifier for SAR Target Discrimination in Complex Scenes

Synthetic aperture radar (SAR) target discrimination is an important stage that distinguishes targets from clutters in the radar automatic target recognition field. However, in complex SAR scenes, the performance of some traditional discriminators will degrade. As an effective tool for one-class classification (OCC), the max-margin one-class classifier has attracted much attention for SAR target discrimination, as it can effectively reduce the impact of multiple clutters. However, the performance of the max-margin one-class classifier is very sensitive to the values of kernel parameters. To solve the problem, this paper proposes an adaptive max-margin one-class classifier for SAR target discrimination in complex scenes. In a max-margin one-class classifier with a suitable kernel parameter, the distance between a sample and classification boundary satisfies a certain geometric relationship, i.e., edge samples in input space are transformed to the region in the kernel space close to boundary, while interior samples in input space are transformed to the region in the kernel space far away from boundary. Therefore, we define the information entropy of samples in the kernel space to measure the distance between samples and classification boundary. To automatically obtain the optimal kernel parameter of the max-margin one-class classifier, the edge and interior samples in the input space are first selected, and then the parameter optimization is performed by minimizing information entropy of interior samples and simultaneously maximizing the information entropy of edge samples. Experimental results of the synthetic datasets and measured synthetic aperture radar (SAR) datasets validate the effectiveness of our method.

Prediction of Carbon Monoxide Concentration with Variation of Support Vector Regression Kernel Parameter Value

Human and industrial activities produce air pollutants that can cause a decline in air quality. In urban areas, transportation activities are the main source of air pollution. One of the emitted air pollutants produced by transportation is carbon monoxide (CO). The understanding of CO concentration is crucial since its overabundance beyond a certain limit will have a negative impact on human health and the environment. In this study, the support vector regression (SVR) method was used to predict CO concentration. The purpose of this study was to predict the hourly CO concentration in the Ujung Berung district, Bandung City, West Java, Indonesia with optimal prediction accuracy. An experiment was carried out by modeling the CO concentration with varying kernel parameter values to obtain accurate prediction results. The suitability of the values between error (ɛ), a trade-off constant (C), and variation mismatch (γ) is vital to obtain optimal prediction results. The results showed that the best prediction accuracy value was 97.68% with kernel parameter values ɛ = 0.02, γ = 30, and C = 0.006. These results may lead to proper decision making on environmental issues and can improve air pollution control strategies.

Bearing defect identification via evolutionary algorithm with adaptive wavelet mutation strategy

The vibration signals obtained from the bearings are associated with nonlinear and nonstationary characteristics, which hinder the fault identification of the bearing. A scheme of fault identification is proposed to identify the faults in bearing in this paper. The vibration signal first pre-processed by wavelet packet transform, followed by Hilbert-Huang transform, decomposes the reconstructed signal into the IMFs. The L-moment criterion is used to select a prominent IMF. The prominent IMF is further utilized to extract the features from the processed signal. The dataset prepared from features trains the SVM model. A novel optimization technique viz 'Diversity driven multi-parent evolutionary algorithm with adaptive wavelet mutation' is proposed to optimize the SVM parameters to increase its efficiency. The efficacy of the proposed algorithm is tested against benchmark functions; obtained results proved its effectiveness. The overall accuracy of built SVM is found to be 100% with suitable regularization and kernel parameter values.

A kernel path algorithm for general parametric quadratic programming problem

It is well known that the performance of a kernel method highly depends on the choice of kernel parameter. A kernel path provides a compact representation of all optimal solutions, which can be used to choose the optimal value of kernel parameter along with cross validation (CV) method. However, none of these existing kernel path algorithms provides a unified implementation to various learning problems. To fill this gap, in this paper, we first study a general parametric quadratic programming (PQP) problem that can be instantiated to an extensive number of learning problems. Then we provide a generalized kernel path (GKP) for the general PQP problem. Furthermore, we analyze the iteration complexity and computational complexity of GKP. Extensive experimental results on various benchmark datasets not only confirm the identity of GKP with several existing kernel path algorithms, but also show that our GKP is superior to the existing kernel path algorithms in terms of generalization and robustness.

Marine oil slicks detection using spaceborne and airborne SAR data

Naval transportations have grown exponentially in the past few years’ and the quantity of illegitimate oil discharges in the oceans has developed with the volume of traffic imposing a severe threat to marine lives and ecosystem. Oil spills dampen the capillary wave’s effect hence create a dark signature on satellite image which is often called a dark object/matter that can be easily detected by SAR sensors. This study reviews the potential of quad-pol and hybrid-pol datasets in detecting oil spills. The data used for this study is the UAVSAR quad-pol and RISAT-1 hybrid/compact-pol dataset. These datasets were acquired during an oil spill experiment named the Norwegian Radar oil Spill Experiment (NORSE2015) which was carried out in the North Sea in June 2015. Basic pre-processing and decomposition modeling techniques namely Freeman and Durden, Van-Zyl, Yamaguchi, and Multiple-Component Scattering Model (MCSM) were applied on UAVSAR dataset, while Raney and Compact-Pol decompositions were applied to RISAT-1 dataset to know different surface backscattering behavior shown by the sea and oil patches. It was observed that Van-Zyl decomposition for UAVSAR and Compact-Pol decomposition for the RISAT-1 dataset gave the most suitable outcome based on the separability analysis to distinguish between oil and water. Two classification techniques Wishart Supervised Classifier (WSC) and Support Vector Machine (SVM) classifications were implemented on decomposed scattering elements of Van-Zyl and Compact-Pol. Radial Basis Function (RBF) kernel parameter values in SVM were used to precisely classify the oil patches from seawater. All the applied decomposition experiments provide good separability between water and oil. Later on, WSC and SVM had been used to finally classify both the dataset outputs. WSC had given better results with an accuracy of 83% for UAVSAR and 86% for RISAT-1, with a larger window size by minimizing the misclassification of sea class into oil spill class. Later the paper concludes the scope of further research possibilities for making a robust system for detecting the oil spills, both precisely and effectually.

A Granular GA-SVM Predictor for Big Data in Agricultural Cyber-Physical Systems

The connection of physical agriculture with corresponding cyber systems is helpful to achieve precision agriculture. Real-time data from agriculture sensors can provide decision supports to improve the yields and quality of agricultural products, but also bring about challenges one of which is how to mine useful information from these vast amounts of data at acceptable computation costs. To deal with the dimension disaster problem faced by most conventional mining algorithms, in this paper we combine granulation techniques and genetic algorithm (GA) with a support vector machine (SVM) to propose a granular GA-SVM. In the integrated predictor, three granulation methods, that is, Min–Median–Max granulation, Quartile–Median granulation, and fuzzy granulation, are introduced to break down big data in agricultural cyber-physical systems into small-scale granules, and GA is used to find the optimal values of SVM penalty parameter and kernel parameter from the reduced granules. Internet of Things (IoT) data from Luochuan Apple Experimental Demonstration Station in Shaanxi Province, China, verified that the proposed granular GA-SVM predictor is effective to make big data prediction with reduced computation time and equivalent accuracy. Moreover, the predicted environment information could provide guidance for growers achieving precise management of apple planting.

Differential mobility and local variation in infection attack rate.

Infectious disease transmission is an inherently spatial process in which a host’s home location and their social mixing patterns are important, with the mixing of infectious individuals often different to that of susceptible individuals. Although incidence data for humans have traditionally been aggregated into low-resolution data sets, modern representative surveillance systems such as electronic hospital records generate high volume case data with precise home locations. Here, we use a gridded spatial transmission model of arbitrary resolution to investigate the theoretical relationship between population density, differential population movement and local variability in incidence. We show analytically that a uniform local attack rate is typically only possible for individual pixels in the grid if susceptible and infectious individuals move in the same way. Using a population in Guangdong, China, for which a robust quantitative description of movement is available (a travel kernel), and a natural history consistent with pandemic influenza; we show that local cumulative incidence is positively correlated with population density when susceptible individuals are more connected in space than infectious individuals. Conversely, under the less intuitively likely scenario, when infectious individuals are more connected, local cumulative incidence is negatively correlated with population density. The strength and direction of correlation changes sign for other kernel parameter values. We show that simulation models in which it is assumed implicitly that only infectious individuals move are assuming a slightly unusual specific correlation between population density and attack rate. However, we also show that this potential structural bias can be corrected by using the appropriate non-isotropic kernel that maps infectious-only code onto the isotropic dual-mobility kernel. These results describe a precise relationship between the spatio-social mixing of infectious and susceptible individuals and local variability in attack rates. More generally, these results suggest a genuine risk that mechanistic models of high-resolution attack rate data may reach spurious conclusions if the precise implications of spatial force-of-infection assumptions are not first fully characterized, prior to models being fit to data.

A Self-tuning Multiobjective Genetic Algorithm with Application in the SVM Classification

In this paper we suggest the self-tuning multiobjective genetic algorithm (STMGA) based on the NSGA-II. This new algorithm is aimed to improve the SVM classification quality. The quality classification indicators such as overall accuracy, specificity, sensitivity and a number of support vectors represent the objective functions in the STMGA. The ways for realizing the self-tuning of the such STMGA parameters as the crossover probability, the crossover distribution index and the mutation distribution index have been proposed and investigated. The considered STMGA is more flexible in the context of selecting its parameters’ values and allows to refuse from the use of the parameters’ values which are set manually. In the case of the radial basis kernel function used for the SVM classifier development, the STMGA finds the Pareto-front of such parameters values as the regularization parameter value and the Gaussian kernel parameter value which give the best values in the chosen set of the classification quality indicators. The experimental results obtained on the basis of the model and real datasets of loan scoring, medical and technical diagnostics, etc. confirm the efficiency of the proposed STMGA.

Research on parameter selection method for support vector machines

The kernel parameter and penalty parameter C are the main factors that affect the learning performance of the support vector machine. However, there are many deficiencies in the existing kernel parameters and penalty parameters C. These methods do not have high accuracy when it comes to classifying multi-category samples, and even ignore some of the samples to conduct training, which violates the integrity of the experimental data. In contrast, this paper improves the selection method of support vector machine kernel parameters and penalty parameters in two ways. First, it obtains the kernel parameter value by optimizing the maximum separation interval between the samples. Second, it optimizes the generalization ability estimation via the influence of the non-boundary support vector on the stability of the support vector machine. The method takes full account of all the training sample data, which is applicable to most sample types, and has the characteristics of low initialization requirements and high-test accuracy. The paper finally uses multiple sets of UCI sample data sets and facial image recognition to verify the method. The experimental results show that the method is feasible, effective and stable.

Dynamical effects of nonlocal interactions in discrete-time growth-dispersal models with logistic-type nonlinearities

The paper is devoted to the study of discrete time and continuous space models with nonlocal resource competition and periodic boundary conditions. We consider generalizations of logistic and Ricker's equations as intraspecific resource competition models with symmetric nonlocal dispersal and interaction terms. Both interaction and dispersal are modeled using convolution integrals, each of which has a parameter describing the range of nonlocality. It is shown that the spatially homogeneous equilibrium of these models becomes unstable for some kernel functions and parameter values by performing a linear stability analysis. To be able to further analyze the behavior of solutions to the models near the stability boundary, weakly nonlinear analysis, a well-known method for continuous time systems, is employed. We obtain Stuart–Landau type equations and give their parameters in terms of Fourier transforms of the kernels. This analysis allows us to study the change in amplitudes of the solutions with respect to ranges of nonlocalities of two symmetric kernel functions. Our calculations indicate that supercritical bifurcations occur near stability boundary for uniform kernel functions. We also verify these results numerically for both models.

Prediction of asphaltene precipitation using support vector regression tuned with genetic algorithms

Abstract Due to the severe and costly problems caused by asphaltene precipitation in petroleum industry, developing a quick and accurate model, to predict the asphaltene precipitation under different conditions, seems crucial. In this study, a new model, namely genetic algorithm – support vector regression (GA-SVR) is proposed, which is applied to predict the amount of asphaltene precipitation. GA is used to select the best optimal values of SVR parameters and kernel parameter, simultaneously, to increase the generalization performance of the SVR. The GA-SVR model is trained and tested on the experimental data sets reported in literature. The performance of the GA-SVR model is compared with two scaling equation models, using statistical error measures and graphical analyses. The results show that the prediction performance of the proposed model, is highly reliable and satisfactory.

Effect of crystallographic orientations on the hydrogen-induced cracking resistance improvement of API 5L X70 pipeline steel under various thermomechanical processing

In this study a thermomechanical processing was suggested to improve the hydrogen-induced cracking (HIC) resistance in X70 pipeline steel. Proposed thermomechanical treatment produced favorable crystallographic textures and significantly increased HIC resistance which is of great interest to petroleum industry. The results showed that the high angle grain boundaries and Kernel parameter values acted as hydrogen-trapping sites, leading to increased HIC susceptibility. Highest HIC resistance was obtained in sample rolled isothermally at 850°C, due to the high proportion of grains oriented with {110} planes parallel to the normal direction and {111}//ND fibers accompanied by negligible fraction of grains oriented in {001}//ND.

Surface electromyography based muscle fatigue progression analysis using modified B distribution time–frequency features

Abstract In this work, an attempt has been made to analyze the progression of muscle fatigue using surface electromyography (sEMG) signals and modified B distribution (MBD) based time–frequency analysis. For this purpose, signals are recorded from biceps brachii muscles of fifty healthy adult volunteers during dynamic contractions. The recorded signals are preprocessed and then subjected to MBD based time–frequency distribution (TFD). The instantaneous median frequency (IMDF) is extracted from the time–frequency matrix for different values of kernel parameter. The linear regression technique is used to model the temporal variations of IMDF. Correlation coefficient is computed in order to select the appropriate value for kernel parameter of MBD based TFD. Further, extended version of frequency domain features namely instantaneous spectral ratio (InstSPR) at low frequency band (LFB), medium frequency band (MFB) and high frequency band (HFB) are extracted from the time–frequency spectrum. In addition to these features, IMDF and instantaneous mean frequency (IMNF) are also calculated. The least square error based linear regression technique is used to track the slope variations of these features. The results show that MBD based time–frequency spectrum is able to provide the instantaneous variations of frequency components associated with fatiguing contractions. The values of InstSPR at MFB and HFB regions, IMDF and IMNF show a decreasing trend during the progression of muscle fatigue. However, an increasing trend is observed in LFB regions. Further the coefficient of variation is calculated for all the features. It is found that the values of IMDF, IMNF and InstSPR in LFB region have lowest variability across different subjects in comparison with other two features. It appears that this method could be useful in analyzing various neuromuscular activities in normal and abnormal conditions.

An Extreme Learning Machine Based on the Mixed Kernel Function of Triangular Kernel and Generalized Hermite Dirichlet Kernel

According to the characteristics that the kernel function of extreme learning machine (ELM) and its performance have a strong correlation, a novel extreme learning machine based on a generalized triangle Hermitian kernel function was proposed in this paper. First, the generalized triangle Hermitian kernel function was constructed by using the product of triangular kernel and generalized Hermite Dirichlet kernel, and the proposed kernel function was proved as a valid kernel function of extreme learning machine. Then, the learning methodology of the extreme learning machine based on the proposed kernel function was presented. The biggest advantage of the proposed kernel is its kernel parameter values only chosen in the natural numbers, which thus can greatly shorten the computational time of parameter optimization and retain more of its sample data structure information. Experiments were performed on a number of binary classification, multiclassification, and regression datasets from the UCI benchmark repository. The experiment results demonstrated that the robustness and generalization performance of the proposed method are outperformed compared to other extreme learning machines with different kernels. Furthermore, the learning speed of proposed method is faster than support vector machine (SVM) methods.

A Hybridization of Enhanced Artificial Bee Colony-Least Squares Support Vector Machines for Price Forecasting

Problem statement: As the performance of Least Squares Support Vector Machines (LSSVM) is highly rely on its value of regularization parameter, γ and kernel parameter, σ2, man-made approach is clearly not an appropriate solution since it may lead to blindness in certain extent. In addition, this technique is time consuming and unsystematic, which consequently affect the generalization performance of LSSVM. Approach: This study presents an enhanced Artificial Bee Colony (ABC) to automatically optimize the hyper parameters of interest. The enhancement involved modifications that provide better exploitation activity by the bees during searching and prevent premature convergence. Later, the prediction process is accomplished by LSSVM. Results and Conclusion: Empirical results obtained indicated that feasibility of proposed technique showed a satisfactory performance by producing better prediction accuracy as compared to standard ABC-LSSVM and Back Propagation Neural Network.

Kernel machines for epilepsy diagnosis via EEG signal classification: A comparative study

We carry out a systematic assessment on a suite of kernel-based learning machines while coping with the task of epilepsy diagnosis through automatic electroencephalogram (EEG) signal classification. The kernel machines investigated include the standard support vector machine (SVM), the least squares SVM, the Lagrangian SVM, the smooth SVM, the proximal SVM, and the relevance vector machine. An extensive series of experiments was conducted on publicly available data, whose clinical EEG recordings were obtained from five normal subjects and five epileptic patients. The performance levels delivered by the different kernel machines are contrasted in terms of the criteria of predictive accuracy, sensitivity to the kernel function/parameter value, and sensitivity to the type of features extracted from the signal. For this purpose, 26 values for the kernel parameter (radius) of two well-known kernel functions (namely, Gaussian and exponential radial basis functions) were considered as well as 21 types of features extracted from the EEG signal, including statistical values derived from the discrete wavelet transform, Lyapunov exponents, and combinations thereof. We first quantitatively assess the impact of the choice of the wavelet basis on the quality of the features extracted. Four wavelet basis functions were considered in this study. Then, we provide the average accuracy (i.e., cross-validation error) values delivered by 252 kernel machine configurations; in particular, 40%/35% of the best-calibrated models of the standard and least squares SVMs reached 100% accuracy rate for the two kernel functions considered. Moreover, we show the sensitivity profiles exhibited by a large sample of the configurations whereby one can visually inspect their levels of sensitiveness to the type of feature and to the kernel function/parameter value. Overall, the results evidence that all kernel machines are competitive in terms of accuracy, with the standard and least squares SVMs prevailing more consistently. Moreover, the choice of the kernel function and parameter value as well as the choice of the feature extractor are critical decisions to be taken, albeit the choice of the wavelet family seems not to be so relevant. Also, the statistical values calculated over the Lyapunov exponents were good sources of signal representation, but not as informative as their wavelet counterparts. Finally, a typical sensitivity profile has emerged among all types of machines, involving some regions of stability separated by zones of sharp variation, with some kernel parameter values clearly associated with better accuracy rates (zones of optimality).

A Novel hybrid genetic algorithm for kernel function and parameter optimization in support vector regression

This study developed a novel model, HGA-SVR, for type of kernel function and kernel parameter value optimization in support vector regression (SVR), which is then applied to forecast the maximum electrical daily load. A novel hybrid genetic algorithm (HGA) was adapted to search for the optimal type of kernel function and kernel parameter values of SVR to increase the accuracy of SVR. The proposed model was tested at an electricity load forecasting competition announced on the EUNITE network. The results showed that the new HGA-SVR model outperforms the previous models. Specifically, the new HGA-SVR model can successfully identify the optimal type of kernel function and all the optimal values of the parameters of SVR with the lowest prediction error values in electricity load forecasting.