NumPy / SciPy Recipes for Data Science: Eigenvalues/Eigenvectors of Covariance Matrices
- Submitted by
- christianbauckhage on 09 November 2015
In this note, we discuss a potential pitfall in using NumPy or SciPy methods to compute eigen decompositions of covariance matrices and show how to avoid it. In short, we discuss why to use “eigh” instead of “eig” when computing eigenvalues and/or eigenvectors of covariance matrices.