Machine learning: a primer

Abstract

Machine learning extracts general principles from observed examples without explicit instructions. In previous columns we have discussed several unsupervised learning methods-for example, clustering and principal component analysis-as well as supervised learning methods such as regression and classification. This month, we begin a series that delves more deeply into algorithms that learn patterns from data to make inferences. This process is called machine learning (ML), a rapidly developing domain closely related to high-dimensional statistics, data mining, pattern recognition, and artificial intelligence. Such methods fall under the broad umbrella of "knowledge discovery", a computational and quantitative approach to characterize and predict complex phenomena described by many variables.