skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: GAPS IN SUPPORT VECTOR OPTIMIZATION

Abstract

We show that the stopping criteria used in many support vector machine (SVM) algorithms working on the dual can be interpreted as primal optimality bounds which in turn are known to be important for the statistical analysis of SVMs. To this end we revisit the duality theory underlying the derivation of the dual and show that in many interesting cases primal optimality bounds are the same as known dual optimality bounds.

Authors:
 [1];  [1];  [1];  [1]
  1. Los Alamos National Laboratory
Publication Date:
Research Org.:
Los Alamos National Lab. (LANL), Los Alamos, NM (United States)
OSTI Identifier:
985890
Report Number(s):
LA-UR-07-0621
TRN: US201017%%68
DOE Contract Number:
AC52-06NA25396
Resource Type:
Conference
Resource Relation:
Conference: 20TH CONFERENCE ON LEARNING THEORY ; 200706 ; SAN DIEGO
Country of Publication:
United States
Language:
English
Subject:
99; ALGORITHMS; DUALITY; LEARNING; OPTIMIZATION; VECTORS

Citation Formats

STEINWART, INGO, HUSH, DON, SCOVEL, CLINT, and LIST, NICOLAS. GAPS IN SUPPORT VECTOR OPTIMIZATION. United States: N. p., 2007. Web.
STEINWART, INGO, HUSH, DON, SCOVEL, CLINT, & LIST, NICOLAS. GAPS IN SUPPORT VECTOR OPTIMIZATION. United States.
STEINWART, INGO, HUSH, DON, SCOVEL, CLINT, and LIST, NICOLAS. Mon . "GAPS IN SUPPORT VECTOR OPTIMIZATION". United States. doi:. https://www.osti.gov/servlets/purl/985890.
@article{osti_985890,
title = {GAPS IN SUPPORT VECTOR OPTIMIZATION},
author = {STEINWART, INGO and HUSH, DON and SCOVEL, CLINT and LIST, NICOLAS},
abstractNote = {We show that the stopping criteria used in many support vector machine (SVM) algorithms working on the dual can be interpreted as primal optimality bounds which in turn are known to be important for the statistical analysis of SVMs. To this end we revisit the duality theory underlying the derivation of the dual and show that in many interesting cases primal optimality bounds are the same as known dual optimality bounds.},
doi = {},
journal = {},
number = ,
volume = ,
place = {United States},
year = {Mon Jan 29 00:00:00 EST 2007},
month = {Mon Jan 29 00:00:00 EST 2007}
}

Conference:
Other availability
Please see Document Availability for additional information on obtaining the full-text document. Library patrons may search WorldCat to identify libraries that hold this conference proceeding.

Save / Share:
  • This work proposes an approach for distribution system load forecasting, which aims to provide highly accurate short-term load forecasting with high resolution utilizing a support vector regression (SVR) based forecaster and a two-step hybrid parameters optimization method. Specifically, because the load profiles in distribution systems contain abrupt deviations, a data normalization is designed as the pretreatment for the collected historical load data. Then an SVR model is trained by the load data to forecast the future load. For better performance of SVR, a two-step hybrid optimization algorithm is proposed to determine the best parameters. In the first step of themore » hybrid optimization algorithm, a designed grid traverse algorithm (GTA) is used to narrow the parameters searching area from a global to local space. In the second step, based on the result of the GTA, particle swarm optimization (PSO) is used to determine the best parameters in the local parameter space. After the best parameters are determined, the SVR model is used to forecast the short-term load deviation in the distribution system.« less
  • Combustion optimization has recently demonstrated its potential to reduce NOx emissions in high capacity coal-fired utility boilers. In the present study, support vector regression (SVR), as well as artificial neural networks (ANN), was proposed to model the relationship between NOx emissions and operating parameters of a 300 MW coal-fired utility boiler. The predicted NOx emissions from the SVR model, by comparing with that of the ANN-based model, showed better agreement with the values obtained in the experimental tests on this boiler operated at different loads and various other operating parameters. The mean modeling error and the correlation factor were 1.58%more » and 0.94, respectively. Then, the combination of the SVR model with ant colony optimization (ACO) to reduce NOx emissions was presented in detail. The experimental results showed that the proposed approach can effectively reduce NOx emissions from the coal-fired utility boiler by about 18.69% (65 ppm). A time period of less than 6 min was required for NOx emissions modeling, and 2 min was required for a run of optimization under a PC system. The computing times are suitable for the online application of the proposed method to actual power plants. 37 refs., 8 figs., 3 tabs.« less
  • This paper proposes an approach for distribution system load forecasting, which aims to provide highly accurate short-term load forecasting with high resolution utilizing a support vector regression (SVR) based forecaster and a two-step hybrid parameters optimization method. Specifically, because the load profiles in distribution systems contain abrupt deviations, a data normalization is designed as the pretreatment for the collected historical load data. Then an SVR model is trained by the load data to forecast the future load. For better performance of SVR, a two-step hybrid optimization algorithm is proposed to determine the best parameters. In the first step of themore » hybrid optimization algorithm, a designed grid traverse algorithm (GTA) is used to narrow the parameters searching area from a global to local space. In the second step, based on the result of the GTA, particle swarm optimization (PSO) is used to determine the best parameters in the local parameter space. After the best parameters are determined, the SVR model is used to forecast the short-term load deviation in the distribution system. The performance of the proposed approach is compared to some classic methods in later sections of the paper.« less
  • Validation of nuclear power reactor signals is often performed by comparing signal prototypes with the actual reactor signals. The signal prototypes are often computed based on empirical data. The implementation of an estimation algorithm which can make predictions on limited data is an important issue. A new machine learning algorithm called support vector machines (SVMS) recently developed by Vladimir Vapnik and his coworkers enables a high level of generalization with finite high-dimensional data. The improved generalization in comparison with standard methods like neural networks is due mainly to the following characteristics of the method. The input data space is transformedmore » into a high-dimensional feature space using a kernel function, and the learning problem is formulated as a convex quadratic programming problem with a unique solution. In this paper the authors have applied the SVM method for data-based state estimation in nuclear power reactors. In particular, they implemented and tested kernels developed at Argonne National Laboratory for the Multivariate State Estimation Technique (MSET), a nonlinear, nonparametric estimation technique with a wide range of applications in nuclear reactors. The methodology has been applied to three data sets from experimental and commercial nuclear power reactor applications. The results are promising. The combination of MSET kernels with the SVM method has better noise reduction and generalization properties than the standard MSET algorithm.« less
  • The authors introduce a method of functionally classifying genes by using gene expression data from DNA microarray hybridization experiments. The method is based on the theory of support vector machines (SVMs). SVMs are considered a supervised computer learning method because they exploit prior knowledge of gene function to identify unknown genes of similar function from expression data. SVMs avoid several problems associated with unsupervised clustering methods, such as hierarchical clustering and self-organizing maps. SVMs have many mathematical features that make them attractive for gene expression analysis, including their flexibility in choosing a similarity function, sparseness of solution when dealing withmore » large data sets, the ability to handle large feature spaces, and the ability to identify outliers. They test several SVMs that use different similarity metrics, as well as some other supervised learning methods, and find that the SVMs best identify sets of genes with a common function using expression data. Finally, they use SVMs to predict functional roles for uncharacterized yeast ORFs based on their expression data.« less