We have just seen what instance-based learning methods are, and we have taken a deep dive into the Nearest Neighbor algorithm and covered specific implementation aspects. In this section, we will look into kernels and the kernel-based Machine learning algorithms.
A kernel, in simple terms, is a similarity function that is fed into a Machine learning algorithm. It takes two inputs and suggests how similar they are. For example, if we are dawned with a task of classifying images, the input data is a key-value pair (image, label). So, in terms of the flow, the image data is taken, features are computed, and a vector of features are fed into the Machine learning algorithm. But, in the case of similarity functions, we can define a kernel function that internally computes the similarity between images, and feed this into the learning algorithm along with the images and label data. The outcome of this is a classifier.
The standard regression or SVM or Perceptron frameworks...