A SVC Iterative Learning Algorithm Based on Sample Selection for Large Samples

• Published in: 2007 International Conference on Machine Learning and Cybernetics Vol. 6; pp. 3308 - 3313
• Main Authors: Zi-Jie Chen, Bo Liu, Xu-Peng He
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.08.2007
• Subjects: Clustering algorithms; Cybernetics; Fuzzy c-means; Iterative algorithm; Iterative algorithms; Iterative methods; Large samples; Large-scale systems; Machine learning; Pharmaceutical technology; Sample selection; Static VAr compensators; Support vector machine; Support vector machine classification; Support vector machines
• ISBN: 1424409721; 9781424409723
• ISSN: 2160-133X
• DOI: 10.1109/ICMLC.2007.4370719

This paper focuses on an effective and efficient support vector machine classification training algorithm for large samples. This method is called 'SVC iterative learning algorithm based on sample selection (short for SVCI)'. Initially, a sample selection strategy based on fuzzy c-means clustering is performed to select partial samples as the first training set, so that common decomposition algorithms are competent and efficient in the small-scale sub-learnings. Furthermore, iterative training is applied to improve the rough learning machine to guarantee performance. Before a new training, another sample selection strategy is carried out to define the new training set. The final optimal classifier is approximate to the one of the original problem. Experiments on several large-scale UCI data sets show that, this iterative algorithm can converge quickly, double training speed and cut down the number of support vectors by a half with losing quite little accuracy.