【1.26 #腾讯会议】Sparse SVM for Sufficient Data Reduction

报告题目:Sparse SVM for Sufficient Data Reduction




报告时间:2020年1月26日 下午4:30-5:30

报告地点:腾讯会议 ID:250 325 425

摘要:Kernel-based methods for support vector machines (SVM) have seen a great advantage in various applications. However, they may incur prohibitive computational costs when the involved sample size is on a large scale. Therefore, data reduction (reducing the number of support vectors) appears to be necessary, which gives rise to the topic of the sparse SVM. Motivated by this, the sparsity constrained kernel SVM optimization is taken into consideration in this paper and is capable of controlling the number of support vectors. Based on the established optimality conditions associated with the stationary equations, a Newton-type method is cast to tackle the sparsity constrained optimization  and turns out to enjoy  the one-step convergence property if the starting point is chosen to be close to a local region of a stationary point, leading to a super-fast computational speed. Numerical comparisons with several powerful solvers demonstrate that the proposed method performs exceptionally well, especially for datasets on large scales, in terms of a much fewer number of support vectors and shorter computational time.