In the previous chapter, we briefly discussed non-parametric machine learning methods. This section will be primarily focused on what the parametric methods of machine learning are, and what they actually learn.
In simple terms, parametric machine learning algorithms try to learn the joint probabilistic distribution of data and their labels. The parameters we learn are of the equation given by joint probabilistic distribution; for example, as we know, logistic regression can be seen as a one-layered neural network. So, considering a one-layered neural network, what it actually learns is the weights and biases of the equation, so as to fit P(Y/X) to the possible categorical distribution of Y(labels).
Logistic regression is a form of discriminative classifier, and in discriminative classifiers, we only focus on P(Y/X), that is, we make an...
