Data Science and Engineering at Enterprise Scale

Chapter 4. Introduction to Machine Learning

Machine learning is a vast subject, and can be daunting at first. Many of its core concepts, however, can be understood with simple examples. These ideas form the basis of our thinking for even the most complex models. We will introduce the basic ideas here (see Table 4-1 for a summary), and will build on them in Chapter 5.

Table 4-1. Basic machine learning terms

Term	Description
Model	The functional form of the prediction. In this chapter, we study the linear model only.
Features	Input variables to the model (x).
Label	Output variable (y). In this chapter, labels are floating-point numbers.
Parameters	The set of numbers ($\mathbf{w}$) that relates the features to the label. Parameters must be discovered from many examples of features and their associated labels.
Linear model	A simple but very useful model where each feature is multiplied by a single parameter. The form of the model is $f(\mathbf{x},\mathbf{w})=w_0+{\sum }_{i=1}^N{w}_i{x}_i$, and is used exclusively in this chapter.
Bias	The parameter in a linear model that does not have a feature associated with it (w0).
Training data	A list of features and associated labels. The data is used to determine the parameters of the model.
Loss function	Also referred to as the objective function. The form of the function to be minimized when determining parameters. The loss function used in this chapter is the least squares function.
Normal equation	A direct solution of the linear model ...