DSPA Chapter 5 Dimensionality Reduction
From Tina Chang
Dimensionality reduction techniques enable exploratory data analyses by reducing the complexity of the dataset, still approximately preserving important properties, such as retaining the distances between cases or subjects. If we are able to reduce the complexity down to a few dimensions, we can then plot the data and untangle its intrinsic characteristics.
We will (1) start with a synthetic example demonstrating the reduction of a 2D data into 1D, (2) explain the notion of rotation matrices, (3) show examples of principal component analysis (PCA), singular value decomposition (SVD), independent component analysis (ICA) and factor analysis (FA), and (4) present a Parkinson’s disease case-study at the end.