Dimensionality Reduction and PCA

Dimensionality Reduction: A Comparative Review

Dimensionality reduction is not just a technical trick. It’s a way of seeing data more clearly, almost like finding the right angle to view a complex sculpture. By reducing dimensions, we reveal the “shape” of the data—clusters, patterns, and relationships that would otherwise be hidden.

Principal component analysis (PCA) is a fancy name for the whole process of reflecting upon what happens along the projections onto the most variable dimensions. It can be used not only for data visualisation and deduplication, but also feature engineering (as in fact it creates new columns that are linear combinations of existing ones).

Dimension Reduction: A Guided Tour (Chapter 3.2)

Linear Dimensionality Reduction: Survey, Insights, and Generalizations