Practical selection of SVM parameters and noise estimation for SVM regression

By Vladimir Cherkassky, Yunqian Ma. Neural Networks, 2004

2,016 citations recorded by the source corpus.

Subfields: Analytical Chemistry, Computer Vision and Pattern Recognition, Statistics and Probability

View Paper: https://www.semanticscholar.org/paper/eaedbadc173e6dda73a0523d763288e27cac56fe

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Analytical Chemistry

In Computer Vision and Pattern Recognition

In Statistics and Probability

Citations in the corpus

A smart antenna system for direction of arrival estimation based on a support vector regression (2005, methodology): “where, following the suggestions in [14], the width parameter is set equal to 0.”
A tutorial review: Metabolomics and partial least squares-discriminant analysis--a marriage of convenience or a shotgun wedding. (2015, methodology): “However, this can be easily optimized by the application of grid search, which identifies the correct values for each of the parameters based on misclassification error [34,112].”
Time of Your Hate: The Challenge of Time in Hate Speech Detection on Social Media (2020, methodology): “A commonly used kernel is RBF (Radial Basis Function) [37] which in general shows good performance for many NLP tasks [38,39].”
Using Data Mining for Wine Quality Assessment (2009, methodology): “To reduce the search space, the first two values will be set using the heuristics [4]: C = 3 (for a standardized output) and = σ̂/ √ N , where σ̂ = 1.”
A Machine Learning Model for Stock Market Prediction (2014, methodology): “SVM compute globally optimal solutions, unlike those obtained with ANNs, which tend to fall into local minima [4].”
Financial time series forecasting using independent component analysis and support vector regression (2009, background): “The performance of SVR is mainly affected by the setting of parameters C and ε [12, 36].”
Using oceanic‐atmospheric oscillations for long lead time streamflow forecasting (2007, methodology): “This is the most widely used approach and has been well documented [Cherkassky and Ma, 2004; Gill et al., 2006; Tokar and Markus, 2000; Asefa et al., 2006; Tripathi et al., 2006].”
Data Mining with Neural Networks and Support Vector Machines Using the R/rminer Tool (2010, methodology): “M=vector("list",3); # vector list of mining M[[1]]=SV;M[[2]]=NN;M[[3]]=MR mgraph(M,graph="REC",leg=c("SVM","NN","MR"),xval=15000,PDF="rec")”
Nondestructive Detection of Soluble Solids Content of Apples from Dielectric Spectra with ANN and Chemometric Methods (2015, methodology): “The SVM regression model can be defined as follows (Cherkassky and Ma 2004): f xð Þ ¼ XN i¼1 αiK x; xið Þ þ b ð8Þ where N is the number of input vectors,αi is the i-th Lagrange multiplier, xi is the i-th input vector, K(x,xi) is the kernel function, and b is the bias.”
Landslide susceptibility mapping using GIS-based multi-criteria decision analysis, support vector machines, and logistic regression (2014, background): “SVR performs linear regression in a high dimension feature space using ε-insensitive loss function and, at the same time, tries to reduce model complexity by minimizing wk k2 (Cherkassky and Ma 2004).”
A survey of remote sensing-based aboveground biomass estimation methods in forest ecosystems (2016, neutral): “The key to this approach is identifying suitable metaparameters: the kernel parameter, precision parameter, and penalty parameter (Cherkassky and Ma 2004).”
Classification of Hyperspectral Images With Regularized Linear Discriminant Analysis (2009, background): “Despite the fact that several criteria are available for selecting the supportvector-machine (SVM) free parameters (see [9] and the references therein), the common approach consists of the exhaustive search through cross-validation strategies (e.”
Stochastic and genetic neural network combinations in trading and hybrid time-varying leverage effects (2014, background): “Consequently the introduction of the parameter C satisfies the need to trade model complexity for training error and vice versa (Cherkassky and Ma, 2004).”