Machine Learning for Industrial Applications

4Kernel Machines

4.1 Different Kernels and Their Applications

In machine learning, kernel machines (KMs) are class pattern analysis algorithms. Their purpose is to find and study the types of relations in datasets. Kernel machines receive their name from the fact that they work in the feature space using kernel functions rather than computing the coordinates of the data in that space. Instead, they compute the inner products between the images of all pairs of data in the feature space. This method is often less computationally demanding than precisely calculating the coordinates. This approach is called the “kernel trick” [1]. Kernel functions have been added for sequence data, text, images, and vectors.

The following are the different types of kernel functions [3, 4]:

1. Polynomial Kernel: The polynomial kernel is a kernel function that represents the similarity of vectors (training samples) in a feature space over polynomials of the original variables, allowing non-linear models to be learned. Support vector machines (SVMs) and other kernelized models are frequently employed with it.
   
2. Sigmoid Kernel: Its primary application is in neural networks. In a two-layer perceptron model of a neural network, this kernel function is analogous to the activation function for neurons.