插值和 SVM 之间的新等效性：核和结构化特征（英文版）

英文标题：New Equivalences Between Interpolation and SVMs: Kernels and Structured Features中文摘要：该论文介绍了一种新的和灵活的分析框架，可用于证明任意再生核希尔伯特空间中的支持向量机，并且在独立亚高斯特征和一般有界正交系统家族（例如傅里叶特征）中的特征两个方面都表现出支持向量增殖现象，这些实验未能覆盖。英文摘要：The support vector machine (SVM) is a supervised learning algorithm thatfinds a maximum-margin linear classifier, often after mapping the data to ahigh-dimensional feature space via the kernel trick. Recent work hasdemonstrated that in certain sufficiently overparameterized settings, the SVMdecision