Does SVD work for non-square matrices?

65 Views Asked by At

I'm trying to decompose a matrix of keypoints for Tomasi-Kanada factorization

I am not doing this by hand right now simply because there are libraries available for this and I know I will get the wrong answer

I am using Real Statistics Resource Pack for Excel

The Matrix I am decomposing is $W$

When I run SVD_U() on the entire matrix I get # Value

But when I run SVD_U() on just the first 6 keypoints, I get a correct $U$ matrix out,

I tested the SVD of this matrix and they do multiply out correctly to original matrix, but this way I cannot use more than 6 keypoints

Is this a problem with this toolkit, or is there a problem with decomposing a rectangular matrix like this?

Am I just doing something wrong?

Or is there a better toolkit for Excel that I can use? I have Matlab as well if that works better

enter image description here