soegaard / sci

Racket libraries for scientific computing
18 stars 4 forks source link

Matrix norms docs #1

Closed Metaxal closed 4 years ago

Metaxal commented 4 years ago

First, excellent work on the documentation!

The docs' definition of ||A||_2 is actually the Frobenius norm ||A||_F (also the entrywise matrix 2-norm), not the matrix (operator) 2-norm. It is probably best to define ||A||_2 as in WP. The Frobenius norm is still pretty useful to have in any case.

||A||_maxabs should be just ||A||max, and it is indeed a [norm](https://en.wikipedia.org/wiki/Norm(mathematics)) as far as I can tell.

soegaard commented 4 years ago

The text was wrong, so I changed it to say Frobenius norm everywhere (which is what the code computes).

I removed the note on the max norm. I misread the LAPACK documentation:

Note that max(abs(A(i,j))) is not a consistent matrix norm.

Metaxal commented 4 years ago

Cool. I think you should also rename 'max-abs to just 'max in the code too, but I won't insist further if you'd rather not to for some reason.

soegaard commented 4 years ago

Missed those. Now fixed.

Metaxal commented 4 years ago

Excellent, thanks!