Add bisection root-finder in arbitrary dimension #7
Conversation
|
Moved from JuliaIntervals/ValidatedNumerics.jl#262. |
dpsanders
changed the title
|
Any comments, @lbenet? |
Codecov Report
@@ Coverage Diff @@
## master #7 +/- ##
=========================================
- Coverage 91.08% 82.18% -8.9%
=========================================
Files 4 5 +1
Lines 157 174 +17
=========================================
Hits 143 143
- Misses 14 31 +17
Continue to review full report at Codecov.
|
|
If you don't have any problem, I'm planning to merge this @lbenet. |
|
I will review it now; if I don't come back to you before lunch time, simple go ahead.... Sorry. |
examples/complex_roots.jl
Outdated
| """ | ||
| function g(X::IntervalBox) | ||
| x, y = X | ||
| z = x + im*y; |
There was a problem hiding this comment.
Not really transcendental, but I read elsewhere that this is better written as z = complex(x, y) (avoids one multiplication and one addition).
There was a problem hiding this comment.
Actually, it maybe even better as z = complex( X... ) directly.
|
I would like to add It's not the best name, but I can't think of a better one right now, and it's pretty nice that it's so easy to write. |
I agree with you. |
|
LGTM. I'll wait to travis before merging. |
| @@ -0,0 +1,34 @@ | |||
| using ValidatedNumerics, ValidatedNumerics.RootFinding | |||
There was a problem hiding this comment.
not in test/REQUIRE
and this file isn't getting run?
This adds a bisection root finder that works with single or multiple variables.
Currently, this involves modifying the
Roottype.For bisection in particular, this is, in some sense, not necessary, since it cannot prove uniqueness, so any roots are
:unknown.However, for multi-dimensional root finding in general, it is maybe necessary to do so.