【摘要】 通过将海量的样本集合合理地分为数目比较小的几个子集,并在每个子集上分别作回归或逼近,使得训练SVM所需二次规划问题的维数大大降低。这样大大降低了训练SVM的运算量,同时提高了局部逼近和预测的能力,为SVM在回归或预测中的实时应用创造了条件。更多还原

【Abstract】 By dividing the large number of samples into some subsets reasonably,we make the regression and approach in the subsets to greatly decrease the dimensions needed by the quadratic programming problems in SVM training.The method reduces the calculations in SVM training,improves the ability of local approaching and prediction and offers a good conditions for the application of SVM in regression and prediction.更多还原