RFC: Add cond() for sparse matrices #12467
Conversation
|
Great. I'll review later today. |
mfasi
force-pushed
the
cond_and_condest
branch
2 times, most recently
from
4ae92c6 to
f6dd68f
Compare
| # TODO rank | ||
|
|
||
| # cond | ||
| function cond(A::SparseMatrixCSC, p::Real=1) |
There was a problem hiding this comment.
cond defaulting to different norms for sparse vs dense is a little iffy.
And I think normestinv would look a little nicer without the leading and trailing lines of whitespace inside the method - the spacing within the code is fine, but right after function and right before end doesn't serve much purpose.
There was a problem hiding this comment.
@tkelman I agree, this bugs me as well. Still, defaulting to a non implemented feature p=2 would be rather annoying. As an alternative, we could enforce the use of both arguments when A is sparse.
You're right, those blank lines aren't that useful. I'll delete them next time I amend the commit.
There was a problem hiding this comment.
What about using 2 as the default here. Eventually, we should support the 2 norm and until we do, people won't accidentally get different norm than they expect.
mfasi
force-pushed
the
cond_and_condest
branch
3 times, most recently
from
1b9a2a8 to
28a2d01
Compare
|
@andreasnoack Than you, I changed what you suggested and amended the commit. |
| n = max(size(A)...) | ||
| F = factorize(A) | ||
| t = min(t,n) | ||
| if length(t) > 1 || t <= 0 |
There was a problem hiding this comment.
length is not necessary here because t is restricted to be an Integer.
There was a problem hiding this comment.
Sorry, I overlooked it. Fixed and amended.
mfasi
force-pushed
the
cond_and_condest
branch
from
28a2d01 to
4ceb4ea
Compare
|
Greta. I've just confirmed that the return type is uniquely defined for |
RFC: Add cond() for sparse matrices
|
Great. I think this closes JuliaLang/LinearAlgebra.jl#104. |
|
Changes Unknown when pulling 4ceb4ea on mfasi:cond_and_condest into ** on JuliaLang:master**. |
This PR would add the function
cond()for sparse matrices, using the algorithm described in JuliaLang/LinearAlgebra.jl#104.