Accelerating improved twin support vector machine with safe screening rule
- 25 Downloads
Improved twin support vector machine (ITSVM) is a binary classification model with strong theoretical interpretation. Compared with Twin bound support vector machine (TWBSVM), it avoids the matrix inverse operation. However, the disadvantage of slow speed is exposed during the training process. Thus authors are motivated to employ safe screening method to reduce the scale of dual ITSVM and accelerate its computational speed. More specifically, the proposed method is to identify redundant points in advance and eliminate them before actually solving the problem. If safe screening method is directly used to accelerate ITSVM, it will be inevitable to bring matrix inverse operation during the screening process. In this case, a screening method which is distinct from existing safe screening method is devised to avoid calculating inverse matrix. Meanwhile an improved dual coordinate descent method (DCDM) is employed to accelerate ITSVM. Experiments on eleven real data sets are conducted to demonstrate the effectiveness of the proposed acceleration method.
KeywordsImproved twin support vector machine Safe screening rule Binary classification Matrix inverse operation
The authors would like to thank the reviewers for the helpful comments and suggestions, which have improved the presentation. This work was supported in part by National Natural Science Foundation of China (No. 11671010) and Beijing Natural Science Foundation (No.4172035).
- 3.Schölkop B, Smola AJ (2002) Learning with kernels. MIT Press, CambridgeGoogle Scholar
- 8.Brugger D (2006) Parallel support vector machines. Universität Tübingen, TübingenGoogle Scholar
- 22.Ogawa K, Suzuki Y, Takeuchi I (2013) Safe screening of mon-support vectors in pathwise SVM computation. In: International conference on machine learning, pp 1382–1390Google Scholar
- 23.Wang J, Zhou J, Wonka P, Ye J (2013) Lasso screening rules via dual polytope projection. In: Neural information processing systems, pp 1070–1078Google Scholar
- 24.Wang J, Wonka P, Ye J (2014) Scaling svm and least absolute deviations via exact data reduction. In: International conference on machine learning, pp 523–531Google Scholar
- 25.Wang J, Zhou J, Liu J, Wonka P, Ye J (2014) A safe screening rule for sparse logistic regression. In: Neural information processing systems, vol 2, pp 1053–1061Google Scholar
- 26.Yang T, Wang J, Sun Q, Hibar DP, Jahanshad N, Liu L, Wang Y, Zhan L, Thompson PM, Ye J (2015) Detecting genetic risk factors for Alzheimer's disease in whole genome sequence data via lasso screening. In: IEEE international symposium on biomedical imaging, pp 985–989Google Scholar
- 34.Hsieh C, Chang K, Lin C, Keerthi SS, Sundararajan S (2008) A dual coordinate descent method for large-scale linear svm. In: International conference on machine learning, pp 408–415Google Scholar
- 36.Lichman M (2013) UCI machine learning repository. http://archive.ics.uci.edu/ml/index.php