Feasibility pump and updates in GOA - #12
Conversation
bernalde
left a comment
bernalde
left a comment
There was a problem hiding this comment.
Great PR. I just left some minor comments and a question open on how to implement the tabu list from the storms of feasibility pumps (according to the paper it is WAY better than no good cuts, so we might give it a look).
| solve_data.LB = results['Problem'][0]['Lower bound'] | ||
| elif not math.isnan(results['Problem'][0]['Upper bound']): | ||
| solve_data.UB = results['Problem'][0]['Upper bound'] | ||
| # TODO: should we use the bound of the rNLP here? |
There was a problem hiding this comment.
If the solver provides a lower bound, we could use it (in the case for example we do not trust the optimal solution of the rNLP). I think that in general, we decide to trust the bound though.
| TransformationFactory('core.fix_integer_vars').apply_to(fixed_nlp) | ||
|
|
||
| MindtPy.MindtPy_linear_cuts.deactivate() | ||
| # TODO: Do we really need to calculate tmp_duals by default? |
There was a problem hiding this comment.
This is a good point. If we do not add the equality relaxation cuts this is time-consuming and dumb. Consider removing it.
| 0, c_geq*(rhs - value(c.body))) | ||
| except (ValueError, OverflowError) as error: | ||
| except (ValueError, OverflowError, AttributeError) as error: | ||
| # AttributeError is included to condition when value(c.body) belongs to complex type |
There was a problem hiding this comment.
This happens when value(c.body) is a complex number. In some examples, due to the numerical computation, 1+0.0001j may show up.
| config.logger.info("Added %s affine cuts" % counter) | ||
|
|
||
| def add_lazy_nogood_cuts(self, var_values, solve_data, config, opt, feasible=False): | ||
| def add_lazy_no_good_cuts(self, var_values, solve_data, config, opt, feasible=False): |
There was a problem hiding this comment.
I thought there was a TODO above mentioning this. Are we using no-good in single tree? Did you try the taboo list mentioned in the storm of feasibility pump paper?
A way of implementing this is their section 5.2 using https://www.ibm.com/support/knowledgecenter/SSSA5P_12.10.0/ilog.odms.cplex.help/refpythoncplex/html/cplex.callbacks.Context-class.html#reject_candidate
| required_solvers = ('ipopt', 'cplex') | ||
| required_solvers = ('ipopt', 'glpk', 'cplex') | ||
| if all(SolverFactory(s).available() for s in required_solvers): | ||
| subsolvers_available = True |
There was a problem hiding this comment.
This will skip some of the tests of CPLEX is not available although mots of them run with GLPK.
I am a little confused about how they handle an integer solution found by CPLEX at the root node appears in the tabu list. Will the root node make a difference? |
I am pretty sure this had a syntax error (undefined rank variable) and was never used (certainly not covered by any tests).
Convert `Disjunct.indicator_var` from binary to Boolean variable
Allow assigning None to a unitted variable
Update Pyomo book examples & tests for the 3rd edition
pyomo.network / pyomo.gdp interoperability
…o replace-simple
Embed dependencies in setup.py
Move constrained_layout tests to 'expensive' suite
Implementation of LOA single-tree
Update IDAES branch to latest Pyomo master
8 participants
Fixes # .
Summary/Motivation:
Changes proposed in this PR:
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: