-
-
Notifications
You must be signed in to change notification settings - Fork 5.5k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Inverse of a SVD factorization of a complex matrix broken in Julia 1.3 #34866
Comments
Goes without saying that I'll prepare a PR to fix this :) |
Can't you quickly fix it by replacing |
I was thinking |
Isn't it better to take the |
Well, strictly speaking, this would be a slightly different truncation. I chose Personally, I don't mind if we'd drop the truncation altogether. But in any case, I think truncation changes should be discussed in #32880. |
On an unrelated note, shouldn't this issue get a bug label (or similar)? |
But isn't |
Maybe I'm misunderstanding - which might well be the case when talking about floating point numbers - but why would the two be equivalent? Wouldn't it depend on the value of julia> s = rand()
0.9013815392525062
julia> eps(real(typeof(s)))*s == eps(s)
false
julia> eps(real(typeof(s)))*s - eps(s)
8.912460530752368e-17 This value is smaller than |
But if you divide the difference by |
This I get. But isn't it the absolute value that matters here? Eventually, in Sorry for perhaps asking the obvious :) |
On Julia 1.1:
On Julia 1.3:
The relevant
eps(T)
call is here.I guess I'm to blame for this, because it seems this was introduced in #32126. Sorry 🤦♂
The text was updated successfully, but these errors were encountered: