We found a match
Your institution may have rights to this item. Sign in to continue.
- Title
A divide-and-conquer method for large scale <italic>ν</italic>-nonparallel support vector machines.
- Authors
Ju, Xuchan; Tian, Yingjie
- Abstract
Recently, nonparallel support vector machine (NPSVM), a branch of support vector machines (SVMs), is developed and has attracted considerable interest. A kind of developed NPSVM, <italic>ν</italic>-nonparallel support vector machine (<italic>ν</italic>-NPSVM), which inherits the advantages of <italic>ν</italic>-support vector machine (<italic>ν</italic>-SVM) enables us to achieve higher classification accuracy and less time to tune for parameters. However, applications of <italic>ν</italic>-NPSVM to large data sets are seriously hampered by their excessive training time. In this paper, we use divide-and-conquer technique for large scale <italic>ν</italic>-nonparallel support vector machine (DC-<italic>ν</italic>NPSVM) aiming at overcoming this burden. In the division step, we first divide the whole samples into several smaller subsamples and solve smaller subproblems using <italic>ν</italic>-NPSVM independently. We theoretically and experimentally show that objective function value, solutions, and support vectors solved by (DC-<italic>ν</italic>NPSVM) are likely to those of the whole <italic>ν</italic>-NPSVM. In the conquer step, subsolutions from subproblems are used as an initial coordinate descent solver which converges to the optimal solution quickly. Moreover, multi-level (DC-<italic>ν</italic>NPSVM) is adopted to balance the accuracy and efficiency. (DC-<italic>ν</italic>NPSVM) can achieve higher accuracy by tuning the parameters in a smaller range and control the number of support vectors efficiently. Quantities of experiments show our (DC-<italic>ν</italic>NPSVM) outperforms state-of-the-art SVM methods for large scale data sets.
- Subjects
SUPPORT vector machines; BIG data; CLASSIFICATION algorithms; STOCHASTIC convergence; MACHINE learning
- Publication
Neural Computing & Applications, 2018, Vol 29, Issue 9, p497
- ISSN
0941-0643
- Publication type
Article
- DOI
10.1007/s00521-016-2574-3