基于最佳逼近点的不变性常识与SVM的融合方法

Incorporating method of invariance and SVM based on best approximate point

不变性常识与支持向量机的融合技术是近年来支持向量机研究的重点之一,将不变性常识融合于学习模型,有助于提高模型的泛化能力。提出了一种新的不变性常识与支持向量机的融合方法,该方法通过最佳逼近点来代表不变性变换形成的轨迹簇将不变性常识融合于SVM。将该方法应用于MNIST手写数字数据库,与经典SVM方法及Virtual SV(VSV)方法的对比实验结果表明,该方法可以提高SVM的泛化能力。

The incorporating method of invariance and support vector machine （SVM） is an important focus for SVM researches in recent years, and it can help to improve the generalization performance of SVM efficiently, A new incorporating approach is presented, which represents the trajectory manifold ofinvariance transformation by the best approximate point. Compared with the traditional SVM and the Virtual SV based on MNIST handwritten digit database, the presented approach greatly improve the generalization performance of SVM.