Pre-code thoughts, benchmark scope and goals #1
stephane-caron
started this conversation in
General
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
-
The benchmark reported in the README only targets a limited number of problems and does not provide a fair comparison between solvers. For instance, solvers with default settings that return less precise solutions faster are advantaged if we only look at out-of-the-box computation time.
We will improve on this in qpsolvers/qpsolvers#70 following the methodology proposed in the OSQP paper and reproduced in proxqp_benchmark.
This thread is here to discuss work items and exchange feedback. Share your thoughts so that we converge on a meaningful bench!
Scope
benchmark/results.md
with summary reported toREADME.md
Metrics
Validation
Solutions can be validated by casting all inequalities to$l \leq C x \leq u$ and checking that, at the primal-dual solution $(x^*, y^*)$ ,
where$\epsilon_{abs}$ is an absolute tolerance parameter. We could check three settings, such as lax ($\epsilon_{abs} = 10^{-3}$ ), medium ($\epsilon_{abs} = 10^{-5}$ ) and high accuracy ($\epsilon_{abs} = 10^{-7}$ ).
See also
Beta Was this translation helpful? Give feedback.
All reactions