反比例光滑支撑向量机

Inverse Proportional Smooth Support Vector Machine

摘要: 为了提高光滑支撑向量机模型的分类性能，文章给出了一种新的光滑函数反比例函数，并利用光滑技术克服了支撑向量机模型的不可微性，通过数学理论上严谨的论证和分析，证明了所给出光滑函数的性质和基于此光滑函数所建立的反比例光滑支撑向量机模型(ISSVM)的收敛性。实验数值表明，反比例光滑支撑向量机比多项式光滑支撑向量机在分类性能上更有优越性。

Abstract: In order to improve the classified performance of smooth support vector machine model, a new smoothed function inverse proportional function is presented, and smoothing technique is used to overcome the non-differentiability of support vector machine model. Through rigorous demonstration and analysis in mathematical theory, the properties of the given smoothing function and the convergence of the inverse proportional smoothing support vector machine (ISSVM) model based on the smoothing function are proved. The experimental results show that the inverse proportional smoothing support vector machine is superior to the polynomial smoothing support vector machine in classified performance.