Support vector machine classifier with truncated pinball loss

Support vector machine (SVM) has attracted great attentions for the last two decades due to its extensive applications, and thus numerous optimization models have been proposed. To distinguish all of them, in this paper, we introduce a new model equipped with an L 0/1 soft-margin loss (dubbed as L 0/1 -SVM) which well captures the nature of the binary classification. Many of the existing convex/non-convex soft-margin losses can be viewed as a surrogate of the L 0/1 soft-margin loss. Despite the discrete nature of L 0/1 , we manage to establish the existence of global minimizer of the new model as well as revealing the relationship among its minimizers and KKT/P-stationary points. These theoretical properties allow us to take advantage of the alternating direction method of multipliers. In addition, the L 0/1 -support vector operator is introduced as a filter to prevent outliers from being support vectors during the training process. Hence, the method is expected to be relatively robust. Finally, numerical experiments demonstrate that our proposed method generates better performance in terms of much shorter computational time with much fewer number of support vectors when against with some other leading methods in areas of SVM. When the data size gets bigger, its advantage becomes more evident.

show abstract

“…The penalty is fixed at κ for t < −κ and is linear otherwise. SVM with such loss ( Tpin -SVM) can be referred in [14].…”

Section: Non-convex Soft-margin Lossesmentioning

confidence: 99%

Support Vector Machine Classifier via Soft-Margin Loss

Wang

Shao

Zhou

et al. 2022

IEEE Trans. Pattern Anal. Mach. Intell.

View full text Add to dashboard Cite

show abstract

“…The basic idea of SVM is to maximize the distance between wo classes. The distance is traditionally determined by the closest point [37]. Some effectiveness has been shown in study on pattern recognition [38].…”

Section: Support Vector Machine (Svm)mentioning

confidence: 99%

A Cardiotocographic Classification using Feature Selection: A comparative Study

Prasetyo

Prastyo²,

Arti³

2021

j.inf.tech.comp.eng

View full text Add to dashboard Cite

Cardiotocography is a series of inspections to determine the health of the fetus in pregnancy. The inspection process is carried out by recording the baby's heart rate information whether in a healthy condition or contrarily. In addition, uterine contractions are also used to determine the health condition of the fetus. Fetal health is classified into 3 conditions namely normal, suspect, and pathological. This paper was performed to compare a classification algorithm for diagnosing the result of the cardiotocographic inspection. An experimental scheme is performed using feature selection and not using it. CFS Subset Evaluation, Info Gain, and Chi-Square are used to select the best feature which correlated to each other. The data set was obtained from the UCI Machine Learning repository available freely. To find out the performance of the classification algorithm, this study uses an evaluation matrix of precision, Recall, F-Measure, MCC, ROC, PRC, and Accuracy. The results showed that all algorithms can provide fairly good classification. However, the combination of the Random Forest algorithm and the Info Gain Feature Selection gives the best results with an accuracy of 93.74%.

show abstract

“…Cover and Hart proposed the first KNN in 1968, which is a classification algorithm. KNN belongs to instance-based learning and is a type of lazy learning; thus, KNN has no explicit learning process or training phase [13]. KNN data sets already have prior classification and feature values and they can be directly processed after receiving new samples.…”

Section: Related Workmentioning

confidence: 99%

A SVM Multi-Class Image Classification Method Based on DE and KNN in Smart City Management

Shu

Cai

2019

IEEE Access

View full text Add to dashboard Cite

When directly operated in an image, good results are always difficult to achieve via conventional methods because they have poor high-dimensional performance. Support vector machine (SVM) is a type of machine learning method with solid foundation that is developed based on traditional statistics. It is also a theory for statistical estimation and predictive learning of objects. This paper optimizes the structure of SVM classification tree with differential evolution (DE) and designs the corresponding DE algorithm to effectively solve the problem of image classification of complex background cases in smart city management systems. In the training process of SVM classification tree, it obtains an optimal two-class classification scheme in every node by means of DE, initially separates the classes that are easy to be separated and then the less easy ones, and finally adaptively generates the best classification tree. The simulation experiment proves that the proposed algorithm is effective when applied to smart city management systems.

show abstract

Support vector machine classifier with truncated pinball loss

Cited by 119 publications

References 11 publications

Support Vector Machine Classifier via Soft-Margin Loss

Support Vector Machine Classifier via Soft-Margin Loss

A Cardiotocographic Classification using Feature Selection: A comparative Study

A SVM Multi-Class Image Classification Method Based on DE and KNN in Smart City Management

Contact Info

Product

Resources

About