A machine learning technique for reducing the number of input dimensions or variables in training data. By reducing the number of dimensions in your data, machine learning systems can learn more quickly and accurately from training data by focusing only on the specific data from which the learning value can be gained. Fewer dimensions means less computing is needed. Fewer dimensions also allows the use of simpler algorithms that aren’t suited with a large number of dimensions. Often, different dimensions in a data set are correlated in some way, and so we can apply a technique to reduce the dimensions by combining or eliminating those dimensions in some way.