SVD can be seen with different glasses:
- You could say that SVD is a data driven generalization of the fourier transform.
- SVD states that any Matrix transformation consists of a rotation, a strech , and another rotation.
- SVD is a decomposition into n Rank 1 independent matrices.
All of these are true and to fully undertastand SVD, we neeed to understand the algebra and "geometry" behind it. The following Repository is a coolection of professor Steve Brunton's SVD classes (UNIVERSITY OF WASHINGTON) and it also includes practical examples from Andrew Ng's machine learning courses (at COURSERA).
