Add Primal/Dual Integral in MindtPy #2285

ZedongPeng · 2022-02-11T06:21:50Z

Summary/Motivation:

Primal integral, dual integral and primal-dual gap integral are performance measures that reward a balance of speed and solution quality for global and local solvers. This PR adds the calculation of primal integral, dual integral and primal-dual gap in MindtPy.

Changes proposed in this PR:

solve_data.LB, solve_data.UB, solve_data.LB_progress and solve_data.UB_progress are all replaced by solve_data.primal_bound, solve_data.dual_bound, solve_data.primal_bound_progress and solve_data.dual_bound_progress. The log has also been modified.

Legal Acknowledgement

By contributing to this software project, I have read the contribution guide and agree to the following terms and conditions for my contribution:

I agree my contributions are submitted under the BSD license.
I represent I am authorized to make the contributions and grant the license. If my employer has rights to intellectual property that includes these contributions, I represent that I have received permission to make contributions and grant the required license on behalf of that employer.

bernalde

Hi Zedong, great first pass. We need to add also tests for the new functionality. Consider a list of primal and dual bound progress and compute the value of the integral as a test.
This could also be the change to add more unit tests to other pieces of MindtPy, especially cut_generation
Great work as always

bernalde · 2022-02-11T19:32:43Z

pyomo/contrib/mindtpy/iterate.py

-                    # TODO determine solve_data.LB, solve_data.UB is inf or -inf.
+                        'Final bound values: Primal Bound: {}  Dual Bound: {}'.
+                        format(solve_data.primal_bound, solve_data.dual_bound))
+                    # TODO determine solve_data.primal_bound, solve_data.dual_bound is inf or -inf.


Is this TODO still active or can we remove it?

I'm not sure. Is there a case where cycling happens and no feasible solution has been found?

No, cycling requires you to go back to the previously found solution.

pyomo/contrib/mindtpy/iterate.py

bernalde · 2022-02-11T19:36:07Z

pyomo/contrib/mindtpy/mip_solve.py

-    else:
-        solve_data.UB_progress.append(solve_data.UB)
+    # TODO: can we remove the following line?
+    # solve_data.dual_bound_progress.append(solve_data.dual_bound)


At the end is the dual_bound_progress complete? If so, we can remove it.

pyomo/contrib/mindtpy/single_tree.py

pyomo/contrib/mindtpy/util.py

codecov · 2022-02-14T16:54:14Z

Codecov Report

Merging #2285 (391fcb9) into main (ea1f36f) will decrease coverage by 0.05%.
The diff coverage is 96.42%.

@@            Coverage Diff             @@
##             main    #2285      +/-   ##
==========================================
- Coverage   85.29%   85.24%   -0.06%     
==========================================
  Files         611      617       +6     
  Lines       76025    77932    +1907     
==========================================
+ Hits        64849    66431    +1582     
- Misses      11176    11501     +325

Flag	Coverage Δ
linux	`82.76% <95.53%> (+0.05%)`	⬆️
osx	`72.97% <4.46%> (+0.07%)`	⬆️
other	`82.71% <95.53%> (+0.05%)`	⬆️
win	`79.85% <95.53%> (+0.06%)`	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
pyomo/contrib/mindtpy/feasibility_pump.py	`80.64% <ø> (ø)`
pyomo/contrib/mindtpy/initialization.py	`79.50% <ø> (ø)`
pyomo/contrib/mindtpy/mip_solve.py	`84.76% <ø> (+0.25%)`	⬆️
pyomo/contrib/mindtpy/nlp_solve.py	`86.98% <66.66%> (+1.01%)`	⬆️
pyomo/contrib/mindtpy/iterate.py	`83.56% <81.81%> (+5.83%)`	⬆️
pyomo/contrib/mindtpy/single_tree.py	`50.15% <93.33%> (+0.31%)`	⬆️
pyomo/contrib/mindtpy/MindtPy.py	`100.00% <100.00%> (ø)`
pyomo/contrib/mindtpy/config_options.py	`93.42% <100.00%> (+8.65%)`	⬆️
pyomo/contrib/mindtpy/util.py	`91.72% <100.00%> (+0.84%)`	⬆️
...o/contrib/interior_point/linalg/mumps_interface.py	`43.95% <0.00%> (-31.05%)`	⬇️
... and 29 more

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update ea1f36f...391fcb9. Read the comment docs.

bernalde

I left behind some comments and we are still missing a little more tests. I'm reviewing this almost real-time but no pressure!

pyomo/contrib/mindtpy/MindtPy.py

pyomo/contrib/mindtpy/single_tree.py

bernalde

Great PR Zedong, thank you for addressing all of my comments. This is ready for review @michaelbynum @jsiirola

ZedongPeng · 2022-02-21T14:09:13Z

I think the code coverage has been improved. Can we call CodeCov to generate the coverage report again?

jsiirola · 2022-02-21T14:39:17Z

@ZedongPeng, it looks like the coverage of MindtPy has improved, and should not hold up this PR. Reporting is a bit problematic at the moment (the Jenkins builds are currently not running correctly - and are not reporting coverage results, so everything is looking worse than it should in practice).

jsiirola

A couple questions, but nothing to prevent merging (given that the tests are passing). Updated coverage information should be available in ~60 minutes.

pyomo/contrib/mindtpy/MindtPy.py

pyomo/contrib/mindtpy/tests/test_mindtpy_global.py

jsiirola · 2022-02-21T16:12:49Z

pyomo/contrib/mindtpy/tests/test_mindtpy_global_lp_nlp.py

 from pyomo.contrib.mcpp import pyomo_mcpp

-required_solvers = ('baron', 'cplex_persistent')
+required_solvers = ('ipopt', 'cplex_persistent')


Do you not need a global solver for these tests?

I think the Pyomo servers don't have baron installed, so these tests are skipped previously, lowering the code coverage. For these selected models, ipopt is enough to find a global optimal solution.

Baron is present on all testing platforms. On GHA it runs in demo mode (because that is a public testing platform). A licensed version of Baron is available through the Jenkins tests (although Jenkins was down for a couple days, it should be back up now).

Although IPOPT is able to converge in the current version to the optimal solution, I think that this is a fragile test as we have no guarantees that with any newer version of the solvers it will keep working. I would be for reverting this back to baron

OK. @ZedongPeng, if you get the global tests updated this afternoon, we should be able to get this merged in tonight in time for tomorrow's release.

All the problems that we are testing are small enough for the demo license of BARON to work with the NLPs. Let's revent this back to baron but not skip if the license is valid (as the demo version should work)

I just updated the global test. Hope it's not too late.

BARON demo mode handles up to 10 constraints and variables and up to 50 nonlinear operations.

I just added the baron license check, since several models exceed this limitation. @bernalde

ZedongPeng added 6 commits February 9, 2022 17:45

change LB/UB to primal bound and dual bound

cbb023f

rename solution_improved to primal_bound_improved

8a8bad5

rename bound_improved to dual_bound_improved

8f45f0b

add get_primal_integral

a0d34c4

fix bug and add dual integral

0126380

Merge branch 'main' into primal-integral

934faa6

bernalde suggested changes Feb 11, 2022

View reviewed changes

jsiirola added the AT: PRE-TEST INSPECTED label Feb 14, 2022

pyomo-autotest removed the AT: PRE-TEST INSPECTED label Feb 14, 2022

blnicho requested a review from michaelbynum February 15, 2022 19:33

ZedongPeng added 5 commits February 16, 2022 12:47

rename bound_tolerance to absolute_bound_tolerance

32f84ef

add docstring and fix typo

4591eec

add log for primal/dual integral

3cdc8a1

versioning

7935295

simplify initial primal/dual bound calculation

c76d98b

bernalde reviewed Feb 17, 2022

View reviewed changes

pyomo/contrib/mindtpy/MindtPy.py Outdated Show resolved Hide resolved

pyomo/contrib/mindtpy/single_tree.py Outdated Show resolved Hide resolved

ZedongPeng added 5 commits February 18, 2022 00:51

remove redundant description

3844f58

fix typo

22529c0

fix primal/dual log

bec7acb

add config.initial_bound_coef

126e541

add primal/dual integral unit test

2c81174

bernalde approved these changes Feb 18, 2022

View reviewed changes

increase code coverage

27a8b9c

jsiirola added the AT: PRE-TEST INSPECTED label Feb 21, 2022

pyomo-autotest removed the AT: PRE-TEST INSPECTED label Feb 21, 2022

jsiirola added the AT: RETEST label Feb 21, 2022

michaelbynum approved these changes Feb 21, 2022

View reviewed changes

jsiirola approved these changes Feb 21, 2022

View reviewed changes

PEP8 format

843ea47

pyomo-autotest removed the AT: RETEST label Feb 21, 2022

jsiirola added the AT: PRE-TEST INSPECTED label Feb 21, 2022

pyomo-autotest removed the AT: PRE-TEST INSPECTED label Feb 21, 2022

ZedongPeng added 2 commits February 22, 2022 10:15

change ipopt to baron

1981d24

remove baron license check

4a8d805

jsiirola added the AT: PRE-TEST INSPECTED label Feb 22, 2022

add baron license check

391fcb9

jsiirola added AT: PRE-TEST INSPECTED and removed AT: PRE-TEST INSPECTED labels Feb 22, 2022

pyomo-autotest removed the AT: PRE-TEST INSPECTED label Feb 22, 2022

jsiirola merged commit 2eca0aa into Pyomo:main Feb 22, 2022

Add Primal/Dual Integral in MindtPy #2285

Add Primal/Dual Integral in MindtPy #2285

Uh oh!

Conversation

ZedongPeng commented Feb 11, 2022

Summary/Motivation:

Changes proposed in this PR:

Legal Acknowledgement

Uh oh!

bernalde left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

codecov bot commented Feb 14, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

bernalde left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

bernalde left a comment

Choose a reason for hiding this comment

Uh oh!

ZedongPeng commented Feb 21, 2022

Uh oh!

jsiirola commented Feb 21, 2022

Uh oh!

jsiirola left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ZedongPeng Feb 22, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

codecov bot commented Feb 14, 2022 •

edited

Loading

ZedongPeng Feb 22, 2022 •

edited

Loading