中文摘要：

核极限学习机（KELM）可使低维空间中线性不可分的数据变得线性可分,增加了ELM算法的鲁棒性,但KELM算法的输入权值参数采用随机初始化,容易导致算法不稳定.为此,本研究提出用粒子群优化算法对KELM中的权值初始参数进行优化、设定,以得到优化的分类器PSO-KELM.由于该算法输出权值求解采用传统的矩阵求逆运算,导致计算复杂,因此再对KELM的输出权值采用Cholesky分解进行优化.经一些标准基因数据集的实验表明,提出的PSO-KELM算法与已有的ELM、KELM、PSO-ELM相比分类精度更高,适用于基因表达数据分类。

英文摘要：

Kernel extreme learning machines （KELM） increase the robustness of extreme learning machines （ELM） by turning linearly non-separable data in low dimensional space into a linearly separable one. However, the internal weighted parameters of KELM are initialized randomly, which causes the algorithm to be unstable. In this paper, we used the particle swam optimization （PSO） algorithm to obtain an optimal set of initial parameters for KELM and thus created an optimal KELM classifier referred to as PSO-KELM. Since a calculation of the output weights adopting the traditional matrix inversion could result in a more complicated algorithm when calculating the output weights, we used the Cholesky decomposition to optimize the output weights of KELM. Experiments on some standard genetic datasets demonstrate that the PSO-KELM has the highest classification accuracy compared to the existing ELM, KELM, PSO-ELM and other similar algorithms.