Cycle-breaking acceleration for support vector regression

Support vector regression (SVR) is a powerful tool in modeling and prediction tasks with widespread application in many areas. The most representative algorithms to train SVR models are Shevade et al.'s Modification 2 and Lin's WSS1 and WSS2 methods in the LIBSVM library. Both are variants of standard SMO in which the updating pairs selected are those that most violate the Karush-Kuhn-Tucker optimality conditions, to which LIBSVM adds a heuristic to improve the decrease in the objective function. In this paper, and after presenting a simple derivation of the updating procedure based on a greedy maximization of the gain in the objective function, we show how cycle-breaking techniques that accelerate the convergence of support vector machines (SVM) in classification can also be applied under this framework, resulting in significantly improved training times for SVR.     

Fuzzy nile-based support vector regression system

In this paper, we design a fuzzy rule-based support vector regression system. The proposed system utilizes the advantages of fuzzy model and support vector regression to extract support vectors to generate fuzzy if-then rules from the training data set. Based on the first-order hnear Tagaki-Sugeno （TS） model, the structure of rules is identified by the support vector regression and then the consequent parameters of rules are tuned by the global least squares method. Our model is applied to the real world regression task. The simulation results gives promising performances in terms of a set of fuzzy hales, which can be easily interpreted by humans.     

Parallel randomized sampling for support vector machine (SVM) and support vector regression (SVR)

A parallel randomized support vector machine (PRSVM) and a parallel randomized support vector regression (PRSVR) algorithm based on a randomized sampling technique are proposed in this paper. The proposed PRSVM and PRSVR have four major advantages over previous methods. (1) We prove that the proposed algorithms achieve an average convergence rate that is so far the fastest bounded convergence rate, among all SVM decomposition training algorithms to the best of our knowledge. The fast average convergence bound is achieved by a unique priority based sampling mechanism. (2) Unlike previous work (Provably fast training algorithm for support vector machines, 2001) the proposed algorithms work for general linear-nonseparable SVM and general non-linear SVR problems. This improvement is achieved by modeling new LP-type problems based on Karush–Kuhn–Tucker optimality conditions. (3) The proposed algorithms are the first parallel version of randomized sampling algorithms for SVM and SVR. Both the analytical convergence bound and the numerical results in a real application show that the proposed algorithm has good scalability. (4) We present demonstrations of the algorithms based on both synthetic data and data obtained from a real word application. Performance comparisons with SVM^light show that the proposed algorithms may be efficiently implemented.     

Exploiting multi-scale support vector regression for image compression

Unlike traditional neural networks that require predefined topology of the network, support vector regression (SVR) approach can model the data within the given level of accuracy with only a small subset of the training data, which are called support vectors (SVs). This property of sparsity has been exploited as the basis for image compression. In this paper, for still image compression, we propose a multi-scale support vector regression (MS-SVR) approach, which can model the images with steep variations and smooth variations very well resulting in good performance. We test our proposed MS-SVR based algorithm on some standard images. The experimental results verify that the proposed MS-SVR achieves better performance than standard SVR. And in a wide range of compression ratio, MS-SVR is very close to JPEG in terms of peak signal-to-noise ratio (PSNR) but exhibits better subjective quality. Furthermore, MS-SVR even outperforms JPEG on both PSNR and subjective quality when the compression ratio is higher enough, for example 25:1 for Lena image. Even when compared with JPEG-2000, the results show greatly similar trend as those in JPEG experiments, except that the compression ratio is a bit higher where our proposed MS-SVR will outperform JPEG-2000.     

A multiple criteria active learning method for support vector regression

This paper presents a novel active learning method developed in the framework of ε-insensitive support vector regression (SVR) for the solution of regression problems with small size initial training data. The proposed active learning method selects iteratively the most informative as well as representative unlabeled samples to be included in the training set by jointly evaluating three criteria: (i) relevancy, (ii) diversity, and (iii) density of samples. All three criteria are implemented according to the SVR properties and are applied in two clustering-based consecutive steps. In the first step, a novel measure to select the most relevant samples that have high probability to be located either outside or on the boundary of the ε-tube of SVR is defined. To this end, initially a clustering method is applied to all unlabeled samples together with the training samples that are inside the ε-tube (those that are not support vectors, i.e., non-SVs); then the clusters with non-SVs are eliminated. The unlabeled samples in the remaining clusters are considered as the most relevant patterns. In the second step, a novel measure to select diverse samples among the relevant patterns from the high density regions in the feature space is defined to better model the SVR learning function. To this end, initially clusters with the highest density of samples are chosen to identify the highest density regions in the feature space. Then, the sample from each selected cluster that is associated with the portion of feature space having the highest density (i.e., the most representative of the underlying distribution of samples contained in the related cluster) is selected to be included in the training set. In this way diverse samples taken from high density regions are efficiently identified. Experimental results obtained on four different data sets show the robustness of the proposed technique particularly when a small-size initial training set are available.     

Combining flat and structured representations for fingerprint classification with recursive neural networks and support vector machines

We present new fingerprint classification algorithms based on two machine learning approaches: support vector machines (SVMs) and recursive neural networks (RNNs). RNNs are trained on a structured representation of the fingerprint image. They are also used to extract a set of distributed features of the fingerprint which can be integrated in the SVM. SVMs are combined with a new error-correcting code scheme. This approach has two main advantages: (a) It can tolerate the presence of ambiguous fingerprint images in the training set and (b) it can effectively identify the most difficult fingerprint images in the test set. By rejecting these images the accuracy of the system improves significantly. We report experiments on the fingerprint database NIST-4. Our best classification accuracy is of 95.6 percent at 20 percent rejection rate and is obtained by training SVMs on both FingerCode and RNN-extracted features. This result indicates the benefit of integrating global and structured representations and suggests that SVMs are a promising approach for fingerprint classification.     

Real-time prediction of order flowtimes using support vector regression

In a make-to-order production system, a due date must be assigned to new orders that arrive dynamically, which requires predicting the order flowtime in real-time. This study develops a support vector regression model for real-time flowtime prediction in multi-resource, multi-product systems. Several combinations of kernel and loss functions are examined, and results indicate that the linear kernel and the ε$\epsilon$-insensitive loss function yield the best generalization performance. The prediction error of the support vector regression model for three different multi-resource systems of varying complexity is compared to that of classic time series models (exponential smoothing and moving average) and to a feedforward artificial neural network. Results show that the support vector regression model has lower flowtime prediction error and is more robust. More accurately predicting flowtime using support vector regression will improve due-date performance and reduce expenses in make-to-order production environments.     

基于支持向量机回归的麻醉药毒性的QSPR研究

采用支持向量机回归(SVR)方法研究了39个麻醉药毒性的定量构效关系,基于留一法交叉验证的结果,模型的相关系数为0.970。结果表明,所建SVR模型的精度高于逆传播人工神经网络(BPANN)、多元线性回归(MLR)和偏最小二乘法(PLS)所得的结果。     

A new learning schema based on support vector for multi-classification

A novel learning schema SVCMR based on support vector is proposed in this paper to address M-class classification issue. It creates a tree-shaped decision frame where M/2 nodes are constructed with the three-separation model as the basic classifier. A class selection rule is defined to ensure basic classifiers be trained in turn on pair of classes with maximum feature distance. Class contours are extracted as data representatives to reduce training set size. Another point is that parameters involved in SVCMR are learned from data neighborhood, which brings adaptation to various datasets and avoids pricy cost spent on searching parameter spaces. Experiments on real datasets demonstrate the performance of SVCMR can be competitive to those state-of-the-art classifiers but with the higher effectiveness than them.     

Parameter-insensitive kernel in extreme learning for non-linear support vector regression

Support vector regression (SVR) is a state-of-the-art method for regression which uses the ε‐sensitive loss and produces sparse models. However, non-linear SVRs are difficult to tune because of the additional kernel parameter. In this paper, a new parameter-insensitive kernel inspired from extreme learning is used for non-linear SVR. Hence, the practitioner has only two meta-parameters to optimise. The proposed approach reduces significantly the computational complexity yet experiments show that it yields performances that are very close from the state-of-the-art. Unlike previous works which rely on Monte-Carlo approximation to estimate the kernel, this work also shows that the proposed kernel has an analytic form which is computationally easier to evaluate.     

Generalization performance of support vector machines and neural networks in runoff modeling

Effective one-day lead runoff prediction is one of the significant aspects of successful water resources management in arid region. For instance, reservoir and hydropower systems call for real-time or on-line site-specific forecasting of the runoff. In this research, we present a new data-driven model called support vector machines (SVMs) based on structural risk minimization principle, which minimizes a bound on a generalized risk (error), as opposed to the empirical risk minimization principle exploited by conventional regression techniques (e.g. ANNs). Thus, this stat-of-the-art methodology for prediction combines excellent generalization property and sparse representation that lead SVMs to be a very promising forecasting method. Further, SVM makes use of a convex quadratic optimization problem; hence, the solution is always unique and globally optimal. To demonstrate the aforementioned forecasting capability of SVM, one-day lead stream flow of Bakhtiyari River in Iran was predicted using the local climate and rainfall data. Moreover, the results were compared with those of ANN and ANN integrated with genetic algorithms (ANN-GA) models. The improvements in root mean squared error (RMSE) and squared correlation coefficient (R²) by SVM over both ANN models indicate that the prediction accuracy of SVM is at least as good as that of those models, yet in some cases actually better, as well as forecasting of high-value discharges.     

The behaviour of the multi-layer perceptron and the support vector regression learning methods in the prediction of NO and NO2 concentrations in Szeged, Hungary

The main aim of this paper is to predict NO and NO₂ concentrations 4 days in advance by comparing two artificial intelligence learning methods, namely, multi-layer perceptron and support vector machines, on two kinds of spatial embedding of the temporal time series. Hourly values of NO and NO₂ concentrations, as well as meteorological variables were recorded in a cross-road monitoring station with heavy traffic in Szeged, in order to build a model for predicting NO and NO₂ concentrations several hours in advance. The prediction of NO and NO₂ concentrations was performed partly on the basis of their past values, and partly on the basis of temperature, humidity and wind speed data. Since NO can be predicted more accurately, its values were considered primarily when forecasting NO₂. Time series prediction can be interpreted in a way that is suitable for artificial intelligence learning. Two effective learning methods, namely, multi-layer perceptron and support vector regression are used to provide efficient non-linear models for NO and NO₂ time series predictions. Multi-layer perceptron is widely used to predict these time series, but support vector regression has not yet been applied for predicting NO and NO₂ concentrations. Three commonly used linear algorithms were considered as references: 1-day persistence, average of several day persistence and linear regression. Based on the good results of the average of several day persistence, a prediction scheme was introduced, which forms weighted averages instead of simple ones. The optimization of these weights was performed with linear regression in linear case and with the learning methods mentioned in non-linear case. Concerning the NO predictions, the non-linear learning methods give significantly better predictions than the reference linear methods. In the case of NO₂, the improvement of the prediction is considerable, however, it is less notable than for NO.     

Automatic nevirapine concentration interpretation system using support vector regression

Follow-up of human immunodeficiency virus (HIV) patients treated with Nevirapine (NVP) is a necessary process to evaluate the drug resistance and the HIV mutation. It is also usually tested by immunochromatographic (IC) strip test. However, it is difficult to estimate the amount of drug the patient gets by visually inspection of color. In this paper, we propose an automatic interpretation system using a commercialized optical scanner. Several IC strips can be placed at any direction as long as they are on the scanner plate. There are three steps in the system, i.e., light intensity normalization, image segmentation and NVP concentration interpretation. We utilized the Support Vector Regression to interpret the NVP concentration. From the results, we found out the performance of the system is promising and better than that of the linear and nonlinear regression.     

Spacebased estimation of moisture transport in marine atmosphere using support vector regression

An improved algorithm is developed based on support vector regression (SVR) to estimate horizonal water vapor transport integrated through the depth of the atmosphere (Θ) over the global ocean from observations of surface wind-stress vector by QuikSCAT, cloud drift wind vector derived from the Multi-angle Imaging SpectroRadiometer (MISR) and geostationary satellites, and precipitable water from the Special Sensor Microwave/Imager (SSM/I). The statistical relation is established between the input parameters (the surface wind stress, the 850 mb wind, the precipitable water, time and location) and the target data (Θ calculated from rawinsondes and reanalysis of numerical weather prediction model). The results are validated with independent daily rawinsonde observations, monthly mean reanalysis data, and through regional water balance. This study clearly demonstrates the improvement of Θ derived from satellite data using SVR over previous data sets based on linear regression and neural network. The SVR methodology reduces both mean bias and standard deviation compared with rawinsonde observations. It agrees better with observations from synoptic to seasonal time scales, and compare more favorably with the reanalysis data on seasonal variations. Only the SVR result can achieve the water balance over South America. The rationale of the advantage by SVR method and the impact of adding the upper level wind will also be discussed.     

Error tolerance based support vector machine for regression

Most existing online algorithms in support vector machines (SVM) can only grow support vectors. This paper proposes an online error tolerance based support vector machine (ET-SVM) which not only grows but also prunes support vectors. Similar to least square support vector machines (LS-SVM), ET-SVM converts the original quadratic program (QP) in standard SVM into a group of easily solved linear equations. Different from LS-SVM, ET-SVM remains support vectors sparse and realizes a compact structure. Thus, ET-SVM can significantly reduce computational time while ensuring satisfactory learning accuracy. Simulation results verify the effectiveness of the newly proposed algorithm.     

Implementing support vector regression with differential evolution to forecast motherboard shipments

In this study, we investigate the forecasting accuracy of motherboard shipments from Taiwan manufacturers. A generalized Bass diffusion model with external variables can provide better forecasting performance. We present a hybrid particle swarm optimization (HPSO) algorithm to improve the parameter estimates of the generalized Bass diffusion model. A support vector regression (SVR) model was recently used successfully to solve forecasting problems. We propose an SVR model with a differential evolution (DE) algorithm to improve forecasting accuracy. We compare our proposed model with the Bass diffusion and generalized Bass diffusion models. The SVR model with a DE algorithm outperforms the other models on both model fit and forecasting accuracy.     

Experimental study on parameter choices in norm-r support vector regression machines with noisy input

In [1], with the evidence framework, the almost inversely linear dependency between the optimal parameter r in norm-r support vector regression machine r-SVR and the Gaussian input noise is theoretically derived. When r takes a non-integer value, r-SVR cannot be easily realized using the classical QP optimization method. This correspondence attempts to achieve two goals: (1) The Newton-decent-method based implementation procedure of r-SVR is presented here; (2) With this procedure, the experimental studies on the dependency between the optimal parameter r in r-SVR and the Gaussian noisy input are given. Our experimental results here confirm the theoretical claim in [1].