基于CCA对LSSVM分类器的稀疏化

为了对分类最小二乘支持向量机实施有效的稀疏化，以提高分类速率，采用分类相关分析算法，按序提取样本核矩阵的全部分类相关成分，并依据样本核矩阵各列与分类相关成分的相关性，对训练集所有个体按分类的重要性排序，进而可选取最重要的部分个体作为支持向量，并将其余非支持向量的信息转移至支持向量，以提高支持向量的分类表达能力.由此构建一种新的稀疏型最小二乘支持向量机CS LSSVM，并将其应用于多个模式分类的实际问题.测试结果表明，CS LSSVM稀疏性很强，且保持了标准LSSVM的分类性能，还可直接适用于多类问题.

CCA sparse least squares support vector machine classifiers

Classification correlative components of the kernel matrix were extracted in an orderly way by the classification correlative analysis(CCA) algorithm to decrease the number of support vectors of least squares support vector machine(LSSVM) and enhance the classification speed of LSSVM.The classification importance order of all training examples was sorted according to the linear relativity between the kernel matrix column and the classification correlative component.Part of the most important training examples were selected as the support vectors of LSSVM.The classification information of non-support vector examples was transferred to support vectors to enhance their classification ability.A new type of sparse LSSVM named CS-LSSVM was proposed and applied to several real life pattern classification problems.Results showed that CS-LSSVM has high sparseness,holds the classification performance of standard LSSVM,and can be applied directly to multi-classification problems.