论文部分内容阅读
在KLasso模型基础之上,引入多核函数与多核参数重新建立的一种更为广发的非线性的多核KLasso模型(MKLasso模型),采用基于梯度Boosting的思想的算法进行求解,并依据人类观察事物的一个基本特征,即人眼位于数据空间较近时能够看清细节,较远时只能够看清整体结构的特性设计了一种模型选择策略,通过实际的3个数据集设计6组试验,来验证该算法的有效性。模拟试验结果表明:MKLasso模型的预测能力明显优于KLasso模型,其预测均方误差提高了10倍;该算法运行高效,抗噪声能力强,在参数选择方面又有一定自己的优势,可以直接选择核参数,算法大大降低了调试与运算时间。
Based on the KLasso model, a more generalized nonlinear multicomponent KLasso model (MKLasso model), which is reestablished by polynomial functions and polynuclear parameters, is introduced. The algorithm based on gradient Boosting is used to solve the problem. Based on the observations of human beings A basic feature is that the human eye can see the details when the data space is relatively close and can only see the characteristics of the whole structure when it is far away. A model selection strategy is designed. Six sets of experiments are designed through the actual three data sets Verify the validity of the algorithm. Simulation results show that the prediction ability of MKLasso model is better than that of KLasso model, and the mean square error of prediction is improved by 10 times. The algorithm is efficient in operation and strong in anti-noise. It has its own advantages in parameter selection and can be directly selected Nuclear parameters, algorithms greatly reduce the debugging and computing time.