Role of Mathematics in Machine Learning

Abstract

Context: Machine learning is forming an algorithm that proposes a model that can learn from data and do predictions. It is used to know why something works and why one model is better than another. Machine learning is a built-in mathematical prerequisite. Machine learning is powered by a combination of statistics, calculus, linear algebra, and probability. Objective: This paper aims to examine the mathematics behind machine learning. Method: Different digital repositories are searched for research articles. Studies that are most relevant to the objective of the paper are selected for analysis. Result: Selected studies show that machine learning models are designed to perform calculation operations like matrix manipulation of the data for which linear algebra is used. The concept of probability is the dominant framework for dealing with uncertainty. Learning algorithms use the concept of probability to create an analysis on a given training dataset for creative prediction. Machine learning is intrinsically data-driven. The initial data acquired from different sources is in the form of junks, to search out the right data, statistical analysis is performed. Calculus helps to find the direction of change. In which direction should the unknown variable be changed such that the prediction is more optimal with minimum error? Conclusion: Findings of this paper are that the mathematical rules of designing models of machine learning are based on concepts of linear algebra, calculus, statistics, and probability. Statistics is widely used for the estimation of the value of a population parameter. Calculus tells us how to learn and optimize models. Linear algebra ensures the feasibility of algorithms to run on massive data sets. Probability predicts the likelihood of the occurrence of an event.