search

tools4ds / DS701-Course-Notes

Repository for DS701 course book and slides
4 stars 1 forks source link

10 - SVD Lecture improvements #143

Open trgardos opened 1 month ago

trgardos commented 1 month ago

For singular values and singular vectors introduction, make the connection to eigenvalues and eigenvectors for symmetric matrices.

Later in the slides, where A=CR notation is used, perhaps relate it back to the A=USV notation better. Also, here rank (k) is used and in the beginning rank r is used.

sladenheim commented 1 month ago

In my PR #145 I adjusted the rank r to be rank k.

I also think based on the first slide we could probably get rid of the figure, since that is assuming a full rank matrix.

I also think we should mention the sizes of the vectors in the first slide too.

sladenheim commented 1 month ago

As for CR factorization, this is a rank revealing factorizating, C and R do not have to be orthogonal nor diagonal. It is not related to the SVD. I just thought it might be helpful to explain that we could form a generic rank k approximation because the point of the SVD is it gives the optimal rank k approximation.

We could always drop this slide. I think we can make the important point about the SVD without it.