论文部分内容阅读
光滑函数能将不光滑模型变为光滑模型,改善支持向量机的回归性能和效率.Lee 等人用一个光滑函数逼近ε-不敏感损失函数的平方,提出ε-不敏感的光滑支持向量回归机模型(ε-SSVR).本文为求ε-不敏感支持向量回归机的新光滑函数,运用插值函数和复合函数的方法,首先求正号函数的光滑逼近,然后将其复合成ε-不敏感损失函数平方的光滑函数,得到一类新的光滑函数.并从理论上证明该类光滑函数的逼近精度比以往的光滑函数高一个数量级.实验结果表明回归效果得到改善,从而为支持向量回归机提供一类新的光滑函数.
Smooth function can change the unsmoothed model into a smooth model and improve the regression performance and efficiency of the support vector machine.Lee et al. Use a smooth function to approximate the square of the ε-insensitive loss function, and put forward the ε-insensitive smooth support vector regression machine Model (ε-SSVR). In this paper, we seek the new smoothing function of ε-insensitive support vector regression machine. Using the interpolation function and the complex function method, we first seek the smooth approximation of the positive sign function and then compound it into ε-insensitive A new kind of smoothing function is obtained, and the approximation precision of the smoothing function is proved to be an order of magnitude higher than that of the previous smoothing function.The experimental results show that the regression effect is improved, and the result of support vector regression Provide a new class of smooth functions.