Kernel Machines: Introduction

Kernels are a type of similarity measures between observed patterns. By exploiting an important mathematical property, they provide new pattern representation and, at the same time, new perspectives to solve many machine learning problems. In this article, we will describe and motivate the main idea of the kernel approach. Notice that, while the usage of kernels is based on well founded theoretical arguments, we will confine our discussion to the main intuitive functional concepts and definitions.