Evaluation of simple performance measures for tuning SVM hyperparameters

By Kaibo Duan, S. Sathiya Keerthi, Aun Neow Poo. Neurocomputing, 2003

712 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Signal Processing

View Paper: https://www.semanticscholar.org/paper/42fa1ce474e3572724eabc7795bc7b2035b0777b

How researchers see this paper

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

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Signal Processing

Citations in the corpus

Learning the Kernel with Hyperkernels (2005, methodology): “For an introduction to SVMs and kernel methods, the reader is referred to numerous tutorials such as Burges (1998) and books such as Schölkopf and Smola (2002).”
Machine Learning and Radiology (2012, background): “A recent approach to address this difficulty is called hyperparameter learning (Bengio, 2000; Duan et al., 2003).”
Supervised classification and mathematical optimization (2013, background): “See [83] for a comparison of the performance of crossvalidation against that of several error bounds.”
Performance evaluation and hyperparameter tuning of statistical and machine-learning models using spatial data (2018, background): “The tuning of SVMs is important and not trivial due to the sensitivity of the hyperparameters across a wide search space (Duan et al., 2003).”
A nested heuristic for parameter tuning in Support Vector Machines (2014, methodology): “Some papers take as surrogate â(·) of the accuracy a(·) a distribution-free, but kernel-specific, bound on the probability of misclassification, see [12, 18, 48].”
Prediction of protein subcellular localization (2006, neutral): “27”
Predicting subcellular localization of proteins for Gram‐negative bacteria by support vector machines based on n‐peptide compositions (2004, methodology): “We use the cross-validation on different parameters for the model selection (Duan et al. 2003).”
Kernel Sliced Inverse Regression with Applications to Classification (2008, background): “Although the choice of optimal kernel and its associated parameters have been investigated by the machine learning community (known as model selection problem), the optimal model parameters are generally domain-speciﬁc (Keerthi and Lin 2003; Duan, Keerthi and Poo 2003).”
Multiple SVM-RFE for gene selection in cancer classification with expression data (2005, background): “For good performance, these parameters must be chosen carefully [15].”
Variable Selection Using SVM-based Criteria (2003, methodology): “These ﬁndings are similar to those obtained for model selection or hyperparameters tuning (Duan et al., 2002).”
A multiple kernel support vector machine scheme for feature selection and rule extraction from gene expression data of cancer tissue (2007, methodology): “For details of the Grid search algorithm, see [25,53].”
Stochastic and genetic neural network combinations in trading and hybrid time-varying leverage effects (2014, background): “As suggested by Duan et al. (2003) keeping the number of folds moderate, i.e. five, offers efficient parameter estimation with constraining substantially computational costs.”
Efficient multiple hyperparameter learning for log-linear models (2007, methodology): “Comparable g en ralization bounds are not generally known for CRFs; even in SVMs, however, generalization bound -based methods empirically do not outperform simpler methods based on optimizing five-fold cr oss-validation error [8].”
Tikhonov, Ivanov and Morozov regularization for support vector machine learning (2016, background): “In this case, the requirement of the SRMprinciple of precisely considering a series of hypothesis spaces of increasing size, for identifying the optimal class of functions, becomes of paramount importance (Vapnik 1998; Duan et al. 2003).”