ABSTRACT
Machine Learning (ML) techniques can be viewed from the perspective of the problems that they solve, the type of model that is built, and also the training philosophy. This chapter describes a variety of problems that ML algorithms solve across domains. Subsequently, ML algorithms will be conceptualized as tools to visualize multi-dimensional data. This viewpoint allows one to understand ML algorithms through the prism of the assumptions that underlie these algorithms; this is useful for a deep understanding of ML algorithms. Classification and function approximation are two types of problems that ML algorithms predominantly solve. Training of ML models can be realized through supervised or unsupervised training paradigms. An interesting aspect of Data Science (DS)/ML is that these ideas cross-cut several disciplines. Unlike traditional disciplines which have very well defined boundaries, DS/ML techniques are being applied in all domains such as engineering, social sciences, natural sciences, and so on.
