ZI round and shifting

[deryaozyurt]

We have implemented ZI Round and Shifting heuristics in HiGHS and added 5 new options: mip_heuristic_run_RINS, mip_heuristic_run_RENS, mip_heuristic_run_rootReducedCost, mip_heuristic_run_ZIRound, mip_heuristic_run_Shifting. The ability to turn off the SubMIP using heuristics, namely RINS, RENS and RootReducedCost seems to be beneficial on some problems. By default ZI Round and Shifting heuristics are turned off to maintain the current HiGHS code behavior.

The heuristics code is added in HighsPrimalHeuristics.cpp with two new member functions of HighsPrimalHeuristics class. A new utility function (getInfeasibleRows) is needed and implemented in HighsMipSolverData.cpp. We also altered the function tryRoundedPoint in HighsPrimalHeuristics.cpp and added the aforementioned options to HighsOptions.h.

For ZI Round, we followed the algorithm description in Chris Wallace’s paper in Journal of Heuristics (Vol. 16, pp. 715-722, 2010) titled “ZI round, a MIP rounding heuristic”.
We mostly followed the Shifting algorithm described in Timo Berthold’s Diplomarbeit titled “Primal Heuristics for Mixed Integer Programs” with several alterations.

Note that MipTimerStart/Stop is not implemented for the new heuristics.

[codecov]

Codecov Report

Attention: Patch coverage is 96.43917% with 12 lines in your changes missing coverage. Please review.

Project coverage is 79.16%. Comparing base (9f8d397) to head (136c472).
Report is 104 commits behind head on latest.

Files with missing lines	Patch %	Lines
highs/mip/HighsPrimalHeuristics.cpp	97.62%	6 Missing ⚠️
highs/lp_data/HighsOptions.h	66.66%	5 Missing ⚠️
highs/mip/HighsMipSolverData.cpp	97.91%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           latest    #2287      +/-   ##
==========================================
+ Coverage   79.10%   79.16%   +0.06%     
==========================================
  Files         343      343              
  Lines       83376    83694     +318     
==========================================
+ Hits        65951    66258     +307     
- Misses      17425    17436      +11     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[jajhall]

Since the MIP solver doesn't conform completely to the naming conventions, and HighsPrimalHeuristics::shifting and HighsPrimalHeuristics::ziRound are new methods owned by @fwesselm, I don't insist on the wholesale changes so that they conform completely to the naming conventions. However, outside these - particularly in places like HighsOptions.h, I feel that the conventions are important.

[fwesselm]

Since the MIP solver doesn't conform completely to the naming conventions, and HighsPrimalHeuristics::shifting and HighsPrimalHeuristics::ziRound are new methods owned by @fwesselm, I don't insist on the wholesale changes so that they conform completely to the naming conventions. However, outside these - particularly in places like HighsOptions.h, I feel that the conventions are important.

@jajhall, thank you for reviewing this. I have tried to address your comments regarding code outside of HighsPrimalHeuristics::shifting and HighsPrimalHeuristics::ziRound.

[jajhall]

I realise, @fwesselm, that I may not have pointed you to the convention that we use (99% of the time) https://google.github.io/styleguide/cppguide.html

[galabovaa]

CI looks good, please ignore the documentation failing, this will fix itself once latest goes into master fot the upcoming release.

[jajhall]

Thanks @deryaozyurt and @fwesselm!