Search

It is vital to have an exclusive modification in semiconductor production process because of meeting differentiated customer demands in dynamic and competitive global minuscule semiconductor technology market and the highly complex fabrication process. In this paper, we propose a control system based on the dynamic mixed-effect least-square support vector regression (LS-SVR) control system for overlay error compensation with stochastic metrology delay to minimize the misalignment of the patterning process. Moreover, for the stability of the control system in the presence of metrology delay and to deal with nonlinearity among the overlay factors, the novel Lyapunov-based kernel function is merged with the LS-SVR controller. The proposed controller’s operation has been validated and implemented by a major semiconductor manufacturer in Taiwan. The experiments are verified that mixed-effect LS-SVR controller has the higher validity and higher efficiency in comparison with the exponentially weighted moving average (EWMA) and threaded EWMA controllers which had been previously implemented at the company or applied in similar studies. Note to Practitioners—Due to high production complexity in semiconductor manufacturing process, a meticulous and intelligent process control is needed to achieve higher throughput and customer satisfaction. Monitoring a complex system is challenging because the process components and variables operate autonomously and interoperate with other manufacturing segments. This paper proposes a novel run-to-run (R2R) control system to compensate the overlay error during the photolithography process that efficiently deals with the high-mixed manufacturing environment and metrology delay. [ABSTRACT FROM AUTHOR]

We present an approach to extend the functionality and the use of kernel machines in image processing applications. We introduce a novel way to design spatial kernel machines with spatial properties and demonstrate how those newly introduced spatial properties enhance the possibilities of the use of kernel machines in image processing applications as a proof of concept. In this paper, we demonstrate four particular extensions: 1) how to model shapes efficiently with spatially computed kernel parameters in a geometrically scalable way; 2) how to visualize the kernel parameters precisely and intuitively on binary 2D shapes; 3) how to construct a one-class classifier from the binary classifier in a straightforward manner without re-training; and 4) how to use the computed kernel parameters for filtering. The existing literature on kernel machines mostly focuses on estimating the optimal kernel parameters via additional cost function(s). In this paper, instead of employing an additional cost function to estimate the kernel-related parameters, we investigate on an analytical solution to predict the actual kernel parameters locally and show how to build a spatial kernel machine with our analytical approach. Classical kernel machines do not perform well on precise shape modeling with a low number of support vectors as demonstrated in this paper. However, we demonstrate and visualize that our analytical approach provides a natural means to relate the kernel parameters to the 2D shapes for sparse shape modeling, where the shape boundary represents the decision boundary. For that, we incorporate the selected kernel function’s geometric properties as an additional constraint into the classifier’s optimization problem by defining an easy-to-explain and intuitive concept: similarity domains. In our experiments, we study and demonstrate how the resulting new kernel machine enhances the capabilities of the classical kernel machines with applications on shape modeling, (geometrically) scaling the non-linear decision boundary at various scales and precise visualization of the kernel parameters in 2D images. [ABSTRACT FROM AUTHOR]

This paper proposes a data-driven small-signal stability constrained optimal power flow (SSSC-OPF) method with high computational efficiency. Instead of repeating the computational expense small-signal stability analysis via differential and algebraic equations during the iterative OPF process, a computationally cheap surrogate constraint for small-signal stability is developed. To reduce the learning difficulty for small-signal stability boundaries, an efficient sample generation strategy is proposed with sampling space compression. This allows us to use the support vector machine (SVM) with a kernel function to derive the explicit data-driven surrogate constraint for small-signal stability. Penalty factor optimization is proposed to compensate for the error caused by SVM. The learned small-signal stability constraint is embedded into the OPF model for generator control. An examination strategy is also developed to avoid the small-signal instability of re-dispatch caused by the error of the data-driven surrogate model. Comparison results with other model-based and data-driven methods on the IEEE 39-bus and 118-bus systems demonstrate the high computational efficiency and economic benefits of the proposed method. [ABSTRACT FROM AUTHOR]

This paper proposes a dynamic analytics method based on the least squares support vector machine with a hybrid kernel to address real-time prediction problems in the converter steelmaking process. The hybrid kernel function is used to enhance the performance of the existing kernels. To improve the model’s accuracy, the internal parameters are optimized by a differential evolution algorithm. In light of the complex mechanisms of the converter steelmaking process, a multistage modeling strategy is designed instead of the traditional single-stage modeling method. Owing to the dynamic nature of the practical production process, great effort has been made to construct a dynamic model that uses the prediction error information based on the static model. The validity of the proposed method is verified through experiments on real-world data collected from a basic oxygen furnace steelmaking process. The results indicate that the proposed method can successfully solve dynamic prediction problems and outperforms other state-of-the-art methods in terms of prediction accuracy. Note to Practitioners—With the development of cyber-physical systems, abundant real-time data have been collected from the converter steelmaking process. These data provide an opportunity to solve product quality prediction problems using data-driven models. This paper proposes a dynamic analytics method based on the least squares support vector machine with a hybrid kernel to address this challenging issue. To improve the model’s performance, a differential evolution algorithm is used to optimize its parameters. Because of the fierce physicochemical reaction in the converter furnace, it is difficult for a single-stage model to achieve accurate predictions. Thus, a multistage modeling strategy is proposed to address this difficulty, and a dynamic model based on feedback error is developed to realize real-time prediction. We verify the effectiveness of the proposed method using real data from a basic oxygen furnace (BOF) steelmaking process. The computational results reveal that the proposed method has a higher prediction accuracy than other methods, making it helpful in guaranteeing the specified product quality and in maintaining stable BOF operation. [ABSTRACT FROM AUTHOR]

Accurate, low-latency, and energy-efficient optical flow estimation is a fundamental kernel function to enable several real-time vision applications on mobile platforms. This paper presents neighbor-guided semi-global matching (NG-fSGM), a new low-complexity optical flow algorithm tailored for low-power mobile applications. NG-fSGM obtains high accuracy optical flow by aggregating local matching costs over a semi-global region, successfully resolving local ambiguity in texture-less and occluded regions. The proposed NG-fSGM aggressively prunes the search space based on neighboring pixels’ information to significantly lower the algorithm complexity from the original fSGM. As a result, NG-fSGM achieves $17.9{\times}$ reduction in the number of computations and $8.37{\times}$ reduction in memory space compared to the original fSGM without compromising its algorithm accuracy. A multicore architecture for NG-fSGM is implemented in hardware to quantify algorithm complexity and power consumption. The proposed architecture realizes NG-fSGM with overlapping blocks processed in parallel to enhance throughput and to lower power consumption. The eight-core architecture achieves 20 M pixel/s (66 frames/s for VGA) throughput with 9.6 mm2 area at 679.2-mW power consumption in 28-nm node. [ABSTRACT FROM AUTHOR]

The prediction accuracy of the traditional kernel with data-driven regression methods strongly depends on the appropriate selection of the kernel function and aims at solving the nonlinearity of the input space. In this paper, a self-learning kernel regression model is proposed. A special kernel space from the measured data is learned and designed, so that combined with dimension reductions on the input variables, the regression behavior between the projected input variables and the output variable is found. The model is posed as a semidefinite programming problem with the objective function to find the maximum variance between the learned manifolds. The kernel is data dependent and can be generated online whenever a new data point is available. The effectiveness of the proposed algorithm is demonstrated through the case studies on a simple nonlinear system and a real semiconductor process. Note to Practitioners—Virtual sensing of quality/key variables is critical for the control and optimization of industrial processes. Traditional data-driven kernel regression methods cannot achieve guaranteed high prediction accuracy. This paper aims to develop a high-accuracy virtual sensor based on the self-learning kernel regression (SLKR) model, which simultaneously constructs the relationship between the input and the output data as well as represents the data in a low-dimensional manifold. For practical implementation, the following are necessary to: 1) select a historical data set with abundant information of the process; 2) preprocess industrial data well and eliminate irregular entries; 3) establish a self-learning regression kernel based on local manifold structure and regression relationship; 4) approximate the explicit data-dependent kernel function and determine suitable parameter; and 5) extract regression coefficients from the learned kernel matrix. [ABSTRACT FROM AUTHOR]

This paper deals with Hadamard \mathbb Z2 \mathbb Z4 {Q}8 -codes, which are binary codes after a Gray map from a subgroup of direct products of \mathbb Z2 , \mathbb Z4 , and Q8 , where Q8 is the quaternionic group. In a previous work, these codes were classified in five shapes. In this paper, we analyze the allowable range of values for the rank and dimension of the kernel, which depends on the particular shape of the code. We show that all these codes can be represented in a standard form, from a set of generators, which can help in understanding the characteristics of each shape. The main results we present are the characterization of Hadamard \mathbb Z2 \mathbb Z4 {Q}8 -codes as a quotient of a semidirect product of \mathbb Z2 \mathbb Z4 -linear codes and the construction of Hadamard \mathbb Z2 \mathbb Z4 {Q}8 -codes with each allowable pair of values for the rank and dimension of the kernel. [ABSTRACT FROM AUTHOR]

Feature template matching (FTM) was proposed by Zhang and Qiu in 2011 for spectrum sensing in cognitive radio. Theoretical analysis for FTM is, however, missing in the literature. This paper will address this issue. Another new direction suggested by this paper is a nonlinear version of FTM, which is called kernel FTM (KFTM). The proposed nonlinear algorithm is performed on data mapped to a kernel space usually with higher dimension by a general nonlinear mapping. The high-dimensional data are cheaply operated with the so-called kernel trick. Furthermore, higher order statistics is considered in the proposed algorithm. Simulations using the real-world measurements of digital television (DTV) signal show that a gain of more than 4 dB can be achieved for KFTM over its linear counterpart FTM. Theoretical analysis is conducted for KFTM (that can be directly applied to FTM). For the first time, a theoretical justification is also performed for FTM. The concentration inequalities of KFTM, which are valid for arbitrary dimensions, are established as a result of concentration of measure phenomenon. The closed-form expressions of the probability distributions (under null hypothesis) are derived for both FTM and KFTM under different kernel functions. The obtained closed-form results agree with simulations. The closed-form expressions of the decision thresholds for a target false-alarm probability are obtained as well. The thresholds are noise power independent; thus, the proposed algorithm overcomes the difficulty of noise uncertainty. [ABSTRACT FROM PUBLISHER]

This paper introduces a novel framework and corresponding methods for sampling and reconstruction of sparse signals in shift-invariant (SI) spaces. We reinterpret the random demodulator, a system that acquires sparse bandlimited signals, as a system for the acquisition of linear combinations of the samples in the SI setting with the box function as the sampling kernel. The sparsity assumption is exploited by the compressive sensing (CS) paradigm for a recovery of the SI samples from a reduced set of measurements. The SI samples are subsequently filtered by a discrete-time correction filter to reconstruct expansion coefficients of the observed signal. Furthermore, we offer a generalization of the proposed framework to other compactly supported sampling kernels that span a wider class of SI spaces. The generalized method embeds the correction filter in the CS optimization problem which directly reconstructs expansion coefficients of the signal. Both approaches recast an inherently continuous-domain inverse problem in a set of finite-dimensional CS problems in an exact way. Finally, we conduct numerical experiments on signals in polynomial B-spline spaces whose expansion coefficients are assumed to be sparse in a certain transform domain. The coefficients can be regarded as parametric models of an underlying continuous-time signal, obtained from a reduced set of measurements. Such continuous signal representations are particularly suitable for signal processing without converting them into samples. [ABSTRACT FROM AUTHOR]

In this paper we study the problem of finding a support of unknown high-dimensional distributions in the presence of labeling information, called Supervised Novelty Detection (SND). The One-Class Support Vector Machine (SVM) is a widely used kernel-based technique to address this problem. However with the latter approach it is difficult to model a mixture of distributions from which the support might be constituted. We address this issue by presenting a new class of SVM-like algorithms which help to approach multi-class classification and novelty detection from a new perspective. We introduce a new coupling term between classes which leverages the problem of finding a good decision boundary while preserving the compactness of a support with the $l_2$ -norm penalty. First we present our optimization objective in the primal and then derive a dual QP formulation of the problem. Next we propose a Least-Squares formulation which results in a linear system which drastically reduces computational costs. Finally we derive a Pegasos-based formulation which can effectively cope with large data sets that cannot be handled by many existing QP solvers. We complete our paper with experiments that validate the usefulness and practical importance of the proposed methods both in classification and novelty detection settings. [ABSTRACT FROM AUTHOR]

In this paper, we discuss nonparametric estimation of the probability density function (PDF) of a univariate random variable. This problem has been the subject of a vast amount of scientific literature in many domains, while statisticians are mainly interested in the analysis of the properties of proposed estimators, and engineers treat the histogram as a ready-to-use tool for a data set analysis. By considering histogram data as a numerical sequence, a simple approach for PDF estimation is presented in this paper. It is based on basic notions related to the reconstruction of a continuous-time signal from a sequence of samples. When estimating continuous PDFs, it is shown that the proposed approach is as accurate as kernel-based estimators, widely adopted in the statistical literature. Conversely, it can provide better accuracy when the PDF to be estimated exhibits a discontinuous behavior. The main statistical properties of the proposed estimators are derived and then verified by simulations related to the common cases of normal and uniform density functions. The obtained results are also used to derive optimal, i.e., minimum integral of the mean square error, estimators. [ABSTRACT FROM AUTHOR]

Early stopping of iterative algorithms is a widely used form of regularization in statistics, commonly used in conjunction with boosting and related gradient-type algorithms. Although consistency results have been established in some settings, such estimators are less well-understood than their analogues based on penalized regularization. In this paper, for a relatively broad class of loss functions and boosting algorithms (including $L^{2}$ -boost, LogitBoost, and AdaBoost, among others), we exhibit a direct connection between the performance of a stopped iterate and the localized Gaussian complexity of the associated function class. This connection allows us to show that the local fixed point analysis of Gaussian or Rademacher complexities, now standard in the analysis of penalized estimators, can be used to derive optimal stopping rules. We derive such stopping rules in detail for various kernel classes and illustrate the correspondence of our theory with practice for Sobolev kernel classes. [ABSTRACT FROM AUTHOR]

Broadband impedance spectroscopy (BIS) provides a new method for cable fault location, however, this method has not been properly used in locating cable fault of cross-bonded (CB) cables due to the geometric discontinuity of impedance spectroscopy measurement loop. This study addresses the possibility of fault location by BIS in CB cables. First, this paper proposes a fault location method by transforming BIS in frequency domain into the location function in space domain with the aid of integral transformation (IT). The kernel function ej2βx is used to obtain fault location function. Second, this paper demonstrates an impedance spectroscopy measurement loop in CB cables. BIS in CB cables is observed having a generalized orthogonality with the kernel function ej2βx. Given this, the IT method is applicable for fault location of CB cables. Finally, this study carries out the location experiment of cable joints on a 110 kV CB cable in cable trench. The experiment shows that BIS in CB cables could be obtained by the loop proposed in this paper. Besides, the results verify the possibility of fault location in CB cables by IT method using BIS. [ABSTRACT FROM AUTHOR]

Accurate grounding parameters calculation ensures the safe and stable operation of the power equipment, and the Green's function calculation of layered soil is the basis of the grounding parameters calculation. The conventional complex image method for Green's function of layered soil mainly include the global equal interval sampling method and the Prony's fitting method, which often have 2 disadvantages: 1. The complex exponential terms of Prony's method are usually twice the number of soil layers, and the amount of calculation increases linearly with the number of soil layers; 2. The global equal interval sampling method cannot guarantee to accurately approximate the integral kernel function, causing the failure to solve Green's function of some complex layered soils. To accurately and efficiently solve the Green's function of arbitrary layered soil, this paper, taking Green's function of horizontally layered soil where the field points and source points are both on the soil surface as an example, summarizes 3 key features of the integral kernel function. Segmented sampling method based on the extreme points of the integral kernel function and the 3-order least squares method are combined to innovatively propose the segmented sampling least squares algorithm (SSLSA) for solving Green's function of layered soil. Only 6 sampling points are required for SSLSA in each sampling segment. The results show that the SSLSA can approximate the integral kernel function more accurately, solve some complicated cases that the conventional complex image method cannot, and has the advantages of good accuracy and efficiency. [ABSTRACT FROM AUTHOR]

In this paper, the noise reduction property of type-2 fuzzy logic (FL) systems (FLSs) (T2FLSs) that use a novel type-2 fuzzy membership function is studied. The proposed type-2 membership function has certain values on both ends of the support and the kernel and some uncertain values for the other values of the support. The parameter tuning rules of a T2FLS that uses such a membership function are derived using the gradient descend learning algorithm. There exist a number of papers in the literature that claim that the performance of T2FLSs is better than type-1 FLSs under noisy conditions, and the claim is tried to be justified by simulation studies only for some specific systems. In this paper, a simpler T2FLS is considered with the novel membership function proposed in which the effect of input noise in the rule base is shown numerically in a general way. The proposed type-2 fuzzy neuro structure is tested on different input–output data sets, and it is shown that the T2FLS with the proposed novel membership function has better noise reduction property when compared to the type-1 counterparts. [ABSTRACT FROM AUTHOR]

When dealing with complex nonlinear signals in intelligent system, by defining the inner product of two data vectors in the feature space, the kernel function can reflect the nonlinear mapping relationship between reproducing kernel Hilbert space (RKHS) and original data space. Therefore, by implementing classical linear adaptive filtering in RKHS space, the filtering operation can be expressed as a special relation of inner product with kernel function, which is referred to as “kernel trick.” As long as these algorithms can be expressed in the form of inner product, not only the convex least squares problem can be solved iteratively, but also the nonlinear adaptive filtering algorithms can be obtained, which have both general approximation characteristics and convexity. Therefore, the combination of kernel method and adaptive filtering algorithm is realized. On the other hand, since the Gram matrix is used, the dimension of kernel adaptive algorithm is determined by the number of data samples. When the number of observation sample point is increasing, the size of the state space increases exponentially with the growth of the dimension. Thereby, the kernel adaptive algorithm should solve the problem of online sparsification to avoid “curse of dimensionality.” As part II, based on online sparse kernel learning and the classical adaptive filtering algorithm, the kernel adaptive algorithms and online sparse algorithms are investigated in this paper. The main works of this paper have two aspects. First, combined with classical adaptive algorithm and kernel feature mapping, this paper investigates the basic concept of kernel adaptive algorithm and the realization mechanism of four kernel adaptive algorithm intensively: 1) kernel least mean squares; 2) kernel recursive least squares; 3) kernel affine projection algorithm; and 4) kernel principal component analysis. Second, in order to reduce the computational complexity, this paper studies some online sparse algorithms which include novel criterion, approximate linear dependency, sliding window, coherence criterion, surprise criterion, and so on. Finally, this paper summarizes the essence of existing conclusions of above algorithms and perspectives the future research direction. [ABSTRACT FROM AUTHOR]

To ensure successful future planetary landing mission, the lander must be capable of detecting and assessing terrain hazard in the nominal landing zone. This paper presents an innovative method of planetary terrain hazard detection. The concept of feature density is introduced into the process of terrain hazard detection. Moreover, the statistical information of the terrain feature density is utilized to represent the terrain hazard level. Opposed to other methods of hazard detection, the research in this paper focuses on how to build a unified terrain representation by the terrain features rather than detecting the specific one or several types of hazard obstacles in traditional sense. Computer simulations demonstrates that the method presented in this paper is able to conduct multitype of complex terrain hazard detecting by the terrain feature detected in the visual image obtained from onboard camera during the landing phase. [ABSTRACT FROM AUTHOR]

It is very important to retrieve a specific person in locating and tracking the missing people as well as the suspects quickly. However, the well-studied face-based and appearance-based individual retrieval methods are ineffective in the surveillance scenarios because of the far photograph distances, the low camera resolutions, the long time intervals, and the complex lighting conditions. To avoid the disadvantages of face-based and appearance-based methods, we propose to retrieve individuals from the surveillance videos with the gait biometric, which has been proved to be beneficial to remote person recognition and robust to lighting variations. What’s more, the gait biometric can be collected without conscious cooperation, making the data collection much easier. But it varies greatly with the view angles, the clothing style, and the carrying conditions. Therefore, the videos of the target person from a similar view angle with the same clothing style and carrying conditions should rank higher than the others. To achieve this purpose and improve the efficiency, this paper proposes a kernel-based semantic hashing (KSH) model, which is learnt by optimizing a semantic triplet ranking loss. Specifically, in the training phase, a semantic similarity score, which depends on the view angles, the clothing style, and the carrying conditions, is calculated for each training pair. Then, a weighted triplet loss considering these semantic scores is designed, which encourages videos with a higher score to stay closer to the gallery in the binary Hamming space. To evaluate the performance of the proposed method, we compare it with several methods on the CASIA Gait Database B and the OU-ISIR Gait Database. The experimental results demonstrate that the KSH is effective and efficient. [ABSTRACT FROM AUTHOR]

The class of random features is one of the most popular techniques to speed up kernel methods in large-scale problems. Related works have been recognized by the NeurIPS Test-of-Time award in 2017 and the ICML Best Paper Finalist in 2019. The body of work on random features has grown rapidly, and hence it is desirable to have a comprehensive overview on this topic explaining the connections among various algorithms and theoretical results. In this survey, we systematically review the work on random features from the past ten years. First, the motivations, characteristics and contributions of representative random features based algorithms are summarized according to their sampling schemes, learning procedures, variance reduction properties and how they exploit training data. Second, we review theoretical results that center around the following key question: how many random features are needed to ensure a high approximation quality or no loss in the empirical/expected risks of the learned estimator. Third, we provide a comprehensive evaluation of popular random features based algorithms on several large-scale benchmark datasets and discuss their approximation quality and prediction performance for classification. Last, we discuss the relationship between random features and modern over-parameterized deep neural networks (DNNs), including the use of high dimensional random features in the analysis of DNNs as well as the gaps between current theoretical and empirical results. This survey may serve as a gentle introduction to this topic, and as a users’ guide for practitioners interested in applying the representative algorithms and understanding theoretical results under various technical assumptions. We hope that this survey will facilitate discussion on the open problems in this topic, and more importantly, shed light on future research directions. Due to the page limit, we suggest the readers refer to the full version of this survey https://arxiv.org/abs/2004.11154. [ABSTRACT FROM AUTHOR]

Renewable energy sources will be more vigorously deployed under the global trend of carbon emission reduction. The connection of numerous renewable energy sources poses an increasingly critical challenge to power quality (PQ) issues in power systems. Time-frequency analysis (TFA) is a foundational technique for real-time monitoring and disturbance detection for power signals. This paper develops an improved S-transform (IST) to accurately detect the disturbances such as oscillatory transient, time-varying harmonics and interharmonics, flicker, swell, sag, interrupt, phase jump, and frequency variation. The proposed IST features the exploration of a designed Gaussian window as the kernel function, whose shape and frequency spectrum can be controlled using a standard deviation based detection frequency parameter. This ensures that the detection requirements at different detection frequencies can be easily met. The IST can be realized by fast Fourier transform (FFT) and its inverse, which ensures that it can be implemented quickly. The IST can accurately detect the amplitude and phase information of fundamental signal, which is beneficial to determine the start and end time, and the intensity of disturbance. With the increase of detection frequency, IST also has excellent energy concentration performance. Simulation and experimental results validated the effectiveness and feasibility of the proposed method. [ABSTRACT FROM AUTHOR]

The accurate assessment of lightning-induced over-voltages is essential for proper insulation coordination studies. In this paper, a novel machine learning based Gaussian process regression model is developed for the prediction of lightning-induced over-voltages considering wide range of ground resistivity and permittivity. The lightning-induced over-voltages have been computed using two-dimensional finite-difference time-domain technique associated with Agrawal field to line coupling model. To account for the input space variability, seven predictor variables namely: resistivity, permittivity, return stroke velocity, distance from flashpoint, return stroke peak current, front-time, and height of an overhead line are considered to determine the response variable, that is, lightning-induced over-voltages. Accordingly, the training of the Gaussian process regression model is carried out using lightning-induced over-voltages obtained from two-dimensional finite-difference time-domain technique for both first and the subsequent strokes. The estimation accuracy of model is appraised using root mean squared error indicating that the exponential kernel function has the least error than other kernel types for describing the covariance. The predictive performance is evaluated by corroborating the predictions from the model against the random test cases generated from two-dimensional finite-difference time-domain technique. The predicted results showed a good agreement with those obtained from finite-difference time-domain technique. [ABSTRACT FROM AUTHOR]

Kernel functions support a broad range of applications that require tasks like density estimation, classification, regression or outlier detection. For these tasks, a common online operation is to compute the weighted aggregation of kernel function values with respect to a set of points. However, scalable aggregation methods are still unknown for typical kernel functions (e.g., Gaussian kernel, polynomial kernel, sigmoid kernel and additive kernels) and weighting schemes. In this paper, we propose a novel and effective bounding technique, by leveraging index structures, to speed up the computation of kernel aggregation. In addition, we extend our technique to additive kernel functions, including $\chi ^2$ χ 2 , intersection, JS and Hellinger kernels, which are widely used in different communities, e.g., computer vision, medical science, Geoscience etc. To handle the additive kernel functions, we further develop the novel and effective bound functions to efficiently evaluate the kernel aggregation. Experimental studies on many real datasets reveal that our proposed solution KARL achieves at least one order of magnitude speedup over the state-of-the-art for different types of kernel functions. [ABSTRACT FROM AUTHOR]

In this paper, we detail effective methods to approximate the achievable rates of channels with additive Gaussian mixture (GM) noise for both real and complex channels to achieve any desired level of accuracy. Attention is paid to a Gaussian input, a discrete real input, and a complex input with discrete amplitude and independent uniform phase. Such discrete inputs represent a wide range of input distributions and they include the capacity-achieving inputs as special cases. At first, we propose a simple technique to accurately calculate the noise entropy. Specifically, when the noise level is high, a lower bound on the integrand of the entropy is established and the noise entropy can be estimated using a closed-form solution. In the low noise region, the piecewise-linear curve fitting (PWLCF) method is applied. We then extend this result to calculate the achievable rate when the input is Gaussian distributed, which is shown to be asymptotically optimal. Next, we propose a simple PWLCF-based method to approximate the output entropy for a real GM channel when the input is discrete, and for a complex GM channel when the input is discrete in amplitude with independent uniform phase. In particular, for the real channel, the output entropy is evaluated by examining the output in high and low regions of amplitude using a lower bound on the integrand of the output entropy and PWLCF, respectively. For the complex channel, the output entropy is approximated a similar manner but using polar coordinates and the Kernel function. It is demonstrated that the output entropy, and consequently, the achievable rates, can be computed to achieve any given accuracy level. [ABSTRACT FROM AUTHOR]

Objective: Pediatric functional nausea is challenging for patients to manage and for clinicians to treat since it lacks objective diagnosis and assessment. A data-driven non-invasive diagnostic screening tool that distinguishes the electro-pathophysiology of pediatric functional nausea from healthy controls would be an invaluable aid to support clinical decision-making in diagnosis and management of patient treatment methodology. The purpose of this paper is to present an innovative approach for objectively classifying pediatric functional nausea using cutaneous high-resolution electrogastrogram data. Methods: We present an Automated Electrogastrogram Data Analytics Pipeline framework and demonstrate its use in a 3x8 factorial design to identify an optimal classification model according to a defined objective function. Low-fidelity synthetic high-resolution electrogastrogram data were generated to validate outputs and determine SOBI-ICA noise reduction effectiveness. Results: A 10 parameter support vector machine binary classifier with a radial basis function kernel was selected as the overall top-performing model from a pool of over 1000 alternatives via maximization of an objective function. This resulted in a 91.6% test ROC AUC score. Conclusion: Using an automated machine learning pipeline approach to process high-resolution electrogastrogram data allows for clinically significant objective classification of pediatric functional nausea. Significance: To our knowledge, this is the first study to demonstrate clinically significant performance in the objective classification of pediatric nausea patients from healthy control subjects using experimental high-resolution electrogastrogram data. These results indicate a promising potential for high-resolution electrogastrography to serve as a data-driven screening tool for the objective diagnosis of pediatric functional nausea. [ABSTRACT FROM AUTHOR]

Learning to improve AUC performance for imbalanced data is an important machine learning research problem. Most methods of AUC maximization assume that the model function is linear in the original feature space. However, this assumption is not suitable for nonlinear separable problems. Although there have been some nonlinear methods of AUC maximization, scaling up nonlinear AUC maximization is still an open question. To address this challenging problem, in this paper, we propose a novel large-scale nonlinear AUC maximization method (named as TSAM) based on the triply stochastic gradient descents. Specifically, we first use the random Fourier feature to approximate the kernel function. After that, we use the triply stochastic gradients w.r.t. the pairwise loss and random feature to iteratively update the solution. Finally, we prove that TSAM converges to the optimal solution with the rate of $ \mathcal {O}(1/t)$ O (1 / t) after $t$ t iterations. Experimental results on a variety of benchmark datasets not only confirm the scalability of TSAM, but also show a significant reduction of computational time compared with existing batch learning algorithms, while retaining the similar generalization performance. [ABSTRACT FROM AUTHOR]

A new, frequency-domain, behavioral modeling methodology for gallium nitride (GaN) high-electron-mobility transistors (HEMTs), based on the Bayesian inference theory, is presented in this paper. Several different probability distribution (kernel) functions are examined for the Bayesian-based modeling architecture, with the optimal kernel function identified through experimental testing. These results are compared to an alternative approach based on the artificial neural networks (ANNs), with the data showing that the proposed approach demonstrates improved accuracy, while at the same time, alleviating the well-known ANN overfitting issue. Model verification is performed at the fundamental and harmonic frequencies using the identified optimal kernel, through comparisons with simulated data from a reference nonlinear circuit model, and with experimental data from separate 2- and 10-W GaN HEMT devices, over a wide range of load conditions. The models can predict accurately the optimal area of the fundamental output power on the Smith chart and the area of optimal power efficiency. Furthermore, the ability of the model to interpolate across input power levels and input frequencies is also tested, with excellent fidelity to the simulated and measured data obtained. [ABSTRACT FROM AUTHOR]

Current clinical dynamic PET has an effective temporal resolution of 5–10 seconds, which can be adequate for traditional compartmental modeling but is inadequate for exploiting the benefit of more advanced tracer kinetic modeling for characterization of diseases (e.g., cancer and heart disease). There is a need to improve dynamic PET to allow fine temporal sampling of 1–2 seconds. However, the reconstruction of these short-time frames from tomographic data is extremely challenging as the count level of each frame is very low and high noise presents in both spatial and temporal domains. Previously, the kernel framework has been developed and demonstrated as a statistically efficient approach to utilizing image prior for low-count PET image reconstruction. Nevertheless, the existing kernel methods mainly explore spatial correlations in the data and only have a limited ability in suppressing temporal noise. In this paper, we propose a new kernel method which extends the previous spatial kernel method to the general spatiotemporal domain. The new kernelized model encodes both spatial and temporal correlations obtained from image prior information and are incorporated into the PET forward projection model to improve themaximumlikelihood(ML) image reconstruction. Computer simulations and an application to real patient scan have shown that the proposed approach can achieve effective noise reduction in both spatial and temporal domains and outperform the spatial kernel method and conventional ML reconstruction method for improving the high temporal-resolution dynamic PET imaging. [ABSTRACT FROM AUTHOR]

Health management for transformer in service is of great significance to the reliable operation of power system. This paper presents a cost-efficient transformer condition prognosis methodology. First, a power management circuit is designed to guarantee the stable performance of self-powered RFID sensor. Moreover, the sensor data storage mechanism is modified by embedding the sensor data into the ID information of RFID sensor to realize less response time and lower power consumption compared with the traditional approach. Second, a denoising method based on band restricted empirical mode decomposition and a fuzzy threshold is introduced to extract noise-free vibration signal from the measured raw signal. Third, the multiple kernel relevance vector machine (MKRVM) is imported to predict the future condition of transformer. The kernels of MKRVM keep diversity for higher prediction accuracy by using the heterogeneous kernel learning approach. Moreover, the double chain quantum genetic algorithm is employed as optimization tool to find the optimal sparse weights of basic kernel functions in MKRVM. The experimental results indicate that the modified RFID sensor can stably operate even when the load of the transformer is abruptly decreased to zero. The response time and power consumption of RFID sensor are markedly decreased by using the improved sensor data storage mechanism. Furthermore, the prediction experiment results demonstrate that the proposed approach is effective in term of health monitoring for transformer. [ABSTRACT FROM AUTHOR]

Visible light communication (VLC) based systems are a viable green supplement to existing radio frequency based communication systems. However, it has been found that the performance of VLC based systems is impaired in conditions when the users are mobile with respect to the transmit luminaire. The relative motion of the mobile users with respect to the luminaire renders the overall VLC channel to be time-varying. Recently, the impact of user mobility on the overall channel impulse response has been modeled by a generalized time-varying VLC channel model, which necessitates for an efficient mechanism at the receiver to tackle this phenomenon. In addition to user mobility, the inter-symbol interference, and the nonlinear characteristics of the light emitting diode are major factors that limit throughput of a VLC-based communication system. To mitigate these impairments, existing techniques such as Volterra/Hammerstein based receivers suffer from modeling error due to truncation of the polynomial kernel till second order terms. Recently, sparse reproducing kernel Hilbert space (RKHS) based methods have been suggested that guarantee universal approximation with the reasonable computational simplicity. However, the choice of a single hyper-parameter restricts its ability to model time-varying channels/systems. Therefore, this paper proposes a novel RKHS based post-distorter that adaptively learns a sparse dictionary based on the incoming observations, and monitors validity of the dictionary based on a proposed metric in RKHS. In order to mitigate the time-varying VLC channel based on this metric, a criterion for clearing the contents of the existing dictionary is proposed, and the requirement to learn a new dictionary is detected. Furthermore, the concept of mixture-adaptive kernel learning is introduced in this work for the minimum symbol error rate (MSER) criterion. From the convergence analysis presented in this paper, faster mean squared error (MSE) convergence is proved for the mixture-kernel based post-distorter. Additionally, it is also proven that for a given step-size, the proposed mixture-kernel MSER post-distorter always converges to a lower MSE as compared to the classical single-kernel MSER. [ABSTRACT FROM AUTHOR]

The conventional calibration-based parallel imaging method assumes a linear relationship between the acquired multi-channel k-space data and the unacquired missing data, where the linear coefficients are estimated using some auto-calibration data. In this paper, we first analyze the model errors in the conventional calibration-based methods and demonstrate the nonlinear relationship. Then, a much more general nonlinear framework is proposed for auto-calibrated parallel imaging. In this framework, kernel tricks are employed to represent the general nonlinear relationship between acquired and unacquired k-space data without increasing the computational complexity. Identification of the nonlinear relationship is still performed by solving linear equations. Experimental results demonstrate that the proposed method can achieve reconstruction quality superior to GRAPPA and NL-GRAPPA at high net reduction factors. [ABSTRACT FROM AUTHOR]

High peak-to-average power ratio (PAPR) of the transmitted signal is a serious issue in multicarrier communication systems, such as second generation terrestrial digital video broadcasting (DVB-T2) systems. These large fluctuations prevent feeding the high power amplifier at an operating point near its non-linear saturation region thereby lowering its power efficiency. In recent years, tone reservation (TR) PAPR reduction techniques have been deeply studied and included in the DVB-T2 and the American digital video broadcasting (ATSC 3.0) specifications. It is based on a gradient iterative approach where, at each iteration, a predefined kernel is used to reduce one peak in time domain. In this paper, a novel TR PAPR reduction technique namely individual carrier allocation for multiple peaks (ICMPs) that is based on a new kernel signal is proposed. This algorithm, compatible with the DVB-T2 standard, offers better performance than the gradient-based DVB-T2 algorithm but suffers with increased complexity for higher modes of DVB-T2 and ATSC 3.0 as the number of required iterations is equal to the number of reserved tones. To overcome this issue, we propose an improved ICMP technique, called grouped ICMP defined with this new kernel. The main principle of this new algorithm consists in dividing the reserved tones into $G$ groups. This highly reduces the number of iterations, now equal to the number of groups, and thereby the latency. An in-depth performance analysis has been done by implementing our algorithm on a testbed platform with real power amplifiers. Both the simulation and experimental results demonstrated that the proposed PAPR reduction algorithm offers very good performance/complexity/latency tradeoff. [ABSTRACT FROM AUTHOR]

Training an accurate and up-to-date radio map has always been a primary concern for implementing a WiFi fingerprint-based localization system. This paper presents a novel radio map learning scheme for online learning a set of kernel density functions, which function like a traditional radio map in WiFi fingerprint-based localization systems. To be specific, each kernel density function corresponds to one specific access point (AP) and predicts the probability of having a received signal strength (RSS) measurement from this AP at any position. In the offline phase, a set of coarse-grained kernel density functions are initially established by the Multivariate Kernel Density Estimation (MKDE) through a fast site survey. In the online phase, these kernel density functions are recursively refined through Incremental-MKDE (IMKDE) using freshly crowdsourced fingerprints, which essentially fuses the historical features in the previous kernel density functions and up-to-date features in the current fingerprints. A compression technique is applied to improve the computation and storage efficiency. In addition, pedestrian dead reckoning (PDR) is leveraged to calibrate the location labels of crowdsourced fingerprints. Extensive experiments are carried out in a real scenario of nearly 1000 $m^2$ for five months, and a comparison made between the IMKDE based method and several existing popular solutions confirms the superiority of the proposed IMKDE in terms of both computational complexity and localization accuracy. [ABSTRACT FROM AUTHOR]

As a new member of convex universal learning machines (CULMs), an augmented space linear model (ASLM) demonstrates strong learning and tracking capabilities in the fields of signal process and machine learning. Unfortunately, the ASLM is only suitable for offline learning, and its estimated accuracy and noise immunity highly depend on the estimated error learned from a generated error table. To address these two issues and apply ASLM into adaptive filtering, online ASLM algorithms with a weighted $k$ nearest neighbor ($k$ -NN) method, based on an updated orthogonal error table, are therefore proposed in this paper. To further improve the robustness of ASLM to heavy-tail noises, a robust ASLM (RASLM) algorithm is proposed with the help of kernel function, and thus a novel robust adaptive filtering algorithm named maximum correntropy criterion RASLM (MCC-RASLM) is developed based the maximum correntropy criterion. However, the linear growth error table existing in MCC-RASLM incurs an ever-growing computational burden. Therefore, to reduce the computational complexity of MCC-RASLM, another novel online variable-centroid clustering is proposed to generate a maximum correntropy criterion cluster RASLM (MCC-C-RASLM) algorithm. Finally, the theoretical analyses regarding robustness and the error bound of RASLM with weighted $k$ -NN regression are performed to verify the effectiveness of MCC-RASLM. Simulation results based on synthetic and real-world data validate the superiorities of the proposed algorithms from aspects of filtering accuracy and robustness. [ABSTRACT FROM AUTHOR]

Linearly derived features have been widely used in hyperspectral image classification to find linear separability of certain classes in recent years. Moreover, nonlinearly transformed features are more effective for class discrimination in real analysis scenarios. However, few efforts have attempted to combine both linear and nonlinear features in the same framework even if they can demonstrate some complementary properties. Moreover, conventional multiple-feature learning-based approaches deal with different features equally, which is not reasonable. This paper proposes an efficient multiple-feature learning-based model with adaptive weights for effectively classifying complex hyperspectral images with limited training samples. A new diversity kernel function is proposed first to simulate the vision perception and analysis procedure of human beings. It could simultaneously evaluate the contrast differences of global features and spatial coherence. Since existing multiple-kernel feature models are always time-consuming, we then design a new adaptive weighted multiple kernel learning method. It employs kernel projection, which could lower the dimensionalities and also learn kernel weights to further discriminate the classification boundaries. For combining both linear and nonlinear features, this paper also proposes a novel decision fusion strategy. The method combines linear and multiple kernel features to balance the classification results of different classifiers. The proposed scheme is tested on several hyperspectral data sets and extended to multisource feature classification environment. The experimental results show that the proposed classification method outperforms most of the existing ones and significantly reduces the computational complexity. [ABSTRACT FROM AUTHOR]

This paper proposes a novel contextual method for classification of polarimetric synthetic aperture radar data. The method combines support vector machine (SVM) and Wishart classifiers to benefit from both parametric and nonparametric methods. This method computes the energy function of a Markov random field (MRF) in the neighborhoods of the pixel using Wishart distribution. It then relates the Markovian energydifference function to the SVM classifier. Therefore, the salt-and-pepper effect on the classified map is reduced using a contextual classifier. Moreover, to achieve the full advantage of spatial information, texture features are added into the contextual classification. Texture features are extracted from SPAN images and are added to the SVM classifier. In this paper, two Radarsat-2 polarimetric images acquired in the leaf-off and leaf-on seasons are used from a forest area. Efficient multitemporal information is exploited using composite kernels in SVM. Comparison of the proposed algorithm with the Wishart, Wishart-MRF, SVM, and SVM with composite kernel classifiers shows a 21.72%, 16.17%, 11.29%, and 8.19% improvement in overall accuracy, respectively. Moreover, incorporating texture features into classification results significant increase in the average accuracy in forest species compared with the use of only polarimetric features. [ABSTRACT FROM PUBLISHER]

Though Fisher discriminant analysis (FDA) is an outstanding method of fault diagnosis, it is usually difficult to extract the discriminant information in a complex industrial environment. One of the reasons is that, in such an environment, the discriminant information can not been extracted entirely due to the disturbances, non-Gaussianity and nonlinearity. In this paper, a method named Joint Fisher discriminant analysis (JFDA) is proposed to address the issues. First, JFDA removes outliers caused by disturbances according to the energy density of each datum. Then, for the non-Gaussianity and weakly nonlinearity, the novel scatter matrices are defined to extract both of the local and global discriminant information based on the manifold learning. Finally, the kernel JFDA (KJFDA) is investigated to hold the manifold assumption because the strongly nonlinearity may weaken the assumption and cause overlapping. The proposed method is applied to the Tennessee Eastman process (TEP). The results demonstrate that KJFDA shows a better performance of fault diagnosis than other improved versions of FDA. [ABSTRACT FROM AUTHOR]