Optim.jl icon indicating copy to clipboard operation
Optim.jl copied to clipboard

Start in interior of variable bounds

Open blegat opened this issue 1 year ago • 5 comments

As highlighted in https://github.com/JuliaNLSolvers/Optim.jl/issues/1071, most tests were failing because the starting point was infeasible. With this PR, most are now passing. The remaining ones to investigate are:

  • test_nonlinear_expression_hs071 : this one gives an infeasible solution, not sure if it's due to the fact that the starting values does not satisfy the constraints, no warning is thrown though.
  • test_nonlinear_expression_hs071_epigraph : the variables are in the interior of the bounds but some of the constraints are not satisfied hence the warning is thrown. Then it converges to an incorrect solution
  • test_objective_FEASIBILITY_SENSE_clears_objective : this one seems it may be a bug in the MOI wrapper, let me investigate first

blegat avatar Jan 29 '24 10:01 blegat

Codecov Report

Attention: Patch coverage is 93.33333% with 1 lines in your changes are missing coverage. Please review.

Project coverage is 85.75%. Comparing base (01b9391) to head (c43736e).

:exclamation: Current head c43736e differs from pull request most recent head 1d9aabb. Consider uploading reports for the commit 1d9aabb to get more accurate results

Files Patch % Lines
src/MOI_wrapper.jl 93.33% 1 Missing :warning:
Additional details and impacted files
@@            Coverage Diff             @@
##           master    #1073      +/-   ##
==========================================
+ Coverage   85.46%   85.75%   +0.29%     
==========================================
  Files          45       46       +1     
  Lines        3274     3511     +237     
==========================================
+ Hits         2798     3011     +213     
- Misses        476      500      +24     

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

codecov[bot] avatar Jan 29 '24 11:01 codecov[bot]

I suggesting investigating the remaining failing tests in separate PRs

blegat avatar Jan 29 '24 13:01 blegat

@pkofod This PR is good to go from my end

blegat avatar Feb 15 '24 10:02 blegat

@pkofod This PR is good to go from my end

thanks I will look. Busy feb and march for me 😓

pkofod avatar Mar 23 '24 00:03 pkofod

Thanks, I just rebased to fix conflicts

blegat avatar Apr 08 '24 08:04 blegat