Adaptative (local) abstol to validate integration
#112
Conversation
Pull Request Test Coverage Report for Build 857850225Warning: This coverage report may be inaccurate.This pull request's base commit is no longer the HEAD commit of its target branch. This means it includes changes from outside the original pull request, including, potentially, unrelated coverage changes.
Details
💛 - Coveralls |
|
This PR aims to shrink the time step in order to validate the integration. It mimics what's currently done in ReachabilityAnalysis.jl. cc @mforets |
|
Nice! I'll try this asap.. |
|
By some reason, some tests (in macos) do not pass, but after a restart of the tests, they do... |
lbenet
force-pushed
the
lb/adaptive_abstol
branch
from
93d4996
to
b3e7ac1
Compare
src/validatedODEs.jl
Outdated
| # If it didn't work and adaptive is false, throw an error | ||
| adaptive && return (false, Δx, t[0]) |
There was a problem hiding this comment.
The comment does not match.
There was a problem hiding this comment.
@schillic if the program is outside the loop it means that "it didn't work", that the program couldn't prove the contraction property. then there are two options: either give up (if adaptive is false), or try again with another tolerance (adaptive is true). the first element in the tuple (false, ...) is a flag to say whether picard iteration succeeded or not.
There was a problem hiding this comment.
Oh sorry, I see now: the error comes below 😅
There was a problem hiding this comment.
Related to this comment: Shall we throw an error (and break the integration) or rather a warning, and (in the main function) stop continuing the integration due to this type of failures?
There was a problem hiding this comment.
(The same would apply in other parts of the code.)
|
@mforets @UzielLinares The last two commits implement ideas we discussed (offline). |
|
Tests pass. Merging! |
No description provided.