Dimensionality reduction
Dimensionality reduction, which is also called feature extraction, refers to the operation to transform a data space given by a large number of dimensions to a subspace of fewer dimensions. The resulting subspace should contain only the most relevant information of the initial data, and the techniques to perform this operation are categorized as linear or non-linear. Dimensionality reduction is a broad class of techniques that is useful for extracting the most relevant information from a large dataset, decreasing its complexity but keeping the relevant information.
The most famous algorithm, Principal Component Analysis (PCA), is a linear mapping of the original data into a subspace of uncorrelated dimensions, and it will be discussed hereafter. The code shown in this paragraph is available in IPython notebook and script versions at the author's GitHub book folder at https://github.com/ai2010/machine_learning_for_the_web/tree/master/chapter_2/.
