def MindtPy_initialize_master(solve_data, config):
"""Initialize the decomposition algorithm.
This includes generating the initial cuts require to build the master
problem.
"""
# if single tree is activated, we need to add bounds for unbounded variables in nonlinear constraints to avoid unbounded master problem.
if config.single_tree:
var_bound_add(solve_data, config)
m = solve_data.mip = solve_data.working_model.clone()
MindtPy = m.MindtPy_utils
m.dual.deactivate()
if config.strategy == 'OA':
calc_jacobians(solve_data, config) # preload jacobians
MindtPy.MindtPy_linear_cuts.oa_cuts = ConstraintList(
doc='Outer approximation cuts')
# elif config.strategy == 'ECP':
# calc_jacobians(solve_data, config) # preload jacobians
# MindtPy.MindtPy_linear_cuts.ecp_cuts = ConstraintList(
# doc='Extended Cutting Planes')
# elif config.strategy == 'PSC':
# detect_nonlinear_vars(solve_data, config)
# MindtPy.MindtPy_linear_cuts.psc_cuts = ConstraintList(
# doc='Partial surrogate cuts')
# elif config.strategy == 'GBD':
# MindtPy.MindtPy_linear_cuts.gbd_cuts = ConstraintList(
# doc='Generalized Benders cuts')
# Set default initialization_strategy
if config.init_strategy is None:
if config.strategy == 'OA':
config.init_strategy = 'rNLP'
else:
config.init_strategy = 'max_binary'
# Do the initialization
elif config.init_strategy == 'rNLP':
init_rNLP(solve_data, config)
elif config.init_strategy == 'max_binary':
init_max_binaries(solve_data, config)
# if config.strategy == 'ECP':
# add_ecp_cut(solve_data, config)
# else:
fixed_nlp, fixed_nlp_result = solve_NLP_subproblem(solve_data, config)
if fixed_nlp_result.solver.termination_condition is tc.optimal or fixed_nlp_result.solver.termination_condition is tc.locallyOptimal:
handle_NLP_subproblem_optimal(fixed_nlp, solve_data, config)
elif fixed_nlp_result.solver.termination_condition is tc.infeasible:
handle_NLP_subproblem_infeasible(fixed_nlp, solve_data, config)
else:
handle_NLP_subproblem_other_termination(fixed_nlp, fixed_nlp_result.solver.termination_condition,
solve_data, config)
def MindtPy_iteration_loop(solve_data, config):
working_model = solve_data.working_model
main_objective = next(
working_model.component_data_objects(Objective, active=True))
while solve_data.mip_iter < config.iteration_limit:
config.logger.info('---MindtPy Master Iteration %s---' %
solve_data.mip_iter)
if algorithm_should_terminate(solve_data, config, check_cycling=False):
break
solve_data.mip_subiter = 0
# solve MILP master problem
if config.strategy == 'OA':
master_mip, master_mip_results = solve_OA_master(
solve_data, config)
if master_mip_results.solver.termination_condition is tc.optimal:
handle_master_mip_optimal(master_mip, solve_data, config)
else:
handle_master_mip_other_conditions(master_mip,
master_mip_results,
solve_data, config)
# Call the MILP post-solve callback
config.call_after_master_solve(master_mip, solve_data)
else:
raise NotImplementedError()
if algorithm_should_terminate(solve_data, config, check_cycling=True):
break
if config.single_tree is False: # if we don't use lazy callback, i.e. LP_NLP
# Solve NLP subproblem
# The constraint linearization happens in the handlers
fixed_nlp, fixed_nlp_result = solve_NLP_subproblem(
solve_data, config)
if fixed_nlp_result.solver.termination_condition is tc.optimal or fixed_nlp_result.solver.termination_condition is tc.locallyOptimal:
handle_NLP_subproblem_optimal(fixed_nlp, solve_data, config)
elif fixed_nlp_result.solver.termination_condition is tc.infeasible:
handle_NLP_subproblem_infeasible(fixed_nlp, solve_data, config)
else:
handle_NLP_subproblem_other_termination(
fixed_nlp, fixed_nlp_result.solver.termination_condition,
solve_data, config)
# Call the NLP post-solve callback
config.call_after_subproblem_solve(fixed_nlp, solve_data)
def MindtPy_iteration_loop(solve_data, config):
working_model = solve_data.working_model
main_objective = next(
working_model.component_data_objects(Objective, active=True))
while solve_data.mip_iter < config.iteration_limit:
config.logger.info('---MindtPy Master Iteration %s---' %
solve_data.mip_iter)
if algorithm_should_terminate(solve_data, config):
break
solve_data.mip_subiter = 0
# solve MILP master problem
if config.strategy == 'OA':
master_mip, master_mip_results = solve_OA_master(
solve_data, config)
if master_mip_results.solver.termination_condition is tc.optimal:
handle_master_mip_optimal(master_mip, solve_data, config)
else:
handle_master_mip_other_conditions(master_mip,
master_mip_results,
solve_data, config)
# Call the MILP post-solve callback
config.call_after_master_solve(master_mip, solve_data)
else:
raise NotImplementedError()
if algorithm_should_terminate(solve_data, config):
break
# Solve NLP subproblem
# The constraint linearization happens in the handlers
fix_nlp, fix_nlp_result = solve_NLP_subproblem(solve_data, config)
if fix_nlp_result.solver.termination_condition is tc.optimal:
handle_NLP_subproblem_optimal(fix_nlp, solve_data, config)
elif fix_nlp_result.solver.termination_condition is tc.infeasible:
handle_NLP_subproblem_infeasible(fix_nlp, solve_data, config)
else:
handle_NLP_subproblem_other_termination(
fix_nlp, fix_nlp_result.solver.termination_condition,
solve_data, config)
# Call the NLP post-solve callback
config.call_after_subproblem_solve(fix_nlp, solve_data)
if config.strategy == 'PSC':
# If the hybrid algorithm is not making progress, switch to OA.
progress_required = 1E-6
if main_objective.sense == minimize:
log = solve_data.LB_progress
sign_adjust = 1
else:
log = solve_data.UB_progress
sign_adjust = -1
# Maximum number of iterations in which the lower (optimistic)
# bound does not improve before switching to OA
max_nonimprove_iter = 5
making_progress = True
# TODO-romeo Unneccesary for OA and LOA, right?
for i in range(1, max_nonimprove_iter + 1):
try:
if (sign_adjust * log[-i] <=
(log[-i - 1] + progress_required) * sign_adjust):
making_progress = False
else:
making_progress = True
break
except IndexError:
# Not enough history yet, keep going.
making_progress = True
break
if not making_progress and (config.strategy == 'hPSC'
or config.strategy == 'PSC'):
config.logger.info(
'Not making enough progress for {} iterations. '
'Switching to OA.'.format(max_nonimprove_iter))
config.strategy = 'OA'
def bound_fix(solve_data, config, last_iter_cuts):
if config.single_tree:
config.logger.info(
'Fix the bound to the value of one iteration before optimal solution is found.'
)
if solve_data.results.problem.sense == ProblemSense.minimize:
solve_data.LB = solve_data.stored_bound[solve_data.UB]
else:
solve_data.UB = solve_data.stored_bound[solve_data.LB]
else:
config.logger.info(
'Solve the master problem without the last nogood cut to fix the bound.'
'zero_tolerance is set to 1E-4')
config.zero_tolerance = 1E-4
# Solve NLP subproblem
# The constraint linearization happens in the handlers
if last_iter_cuts is False:
fixed_nlp, fixed_nlp_result = solve_NLP_subproblem(
solve_data, config)
if fixed_nlp_result.solver.termination_condition in {
tc.optimal, tc.locallyOptimal, tc.feasible
}:
handle_NLP_subproblem_optimal(fixed_nlp, solve_data, config)
elif fixed_nlp_result.solver.termination_condition is tc.infeasible:
handle_NLP_subproblem_infeasible(fixed_nlp, solve_data, config)
else:
handle_NLP_subproblem_other_termination(
fixed_nlp, fixed_nlp_result.solver.termination_condition,
solve_data, config)
MindtPy = solve_data.mip.MindtPy_utils
# only deactivate the last integer cut.
if config.strategy == 'GOA':
if solve_data.results.problem.sense == ProblemSense.minimize:
valid_no_good_cuts_num = solve_data.num_no_good_cuts_added[
solve_data.UB]
else:
valid_no_good_cuts_num = solve_data.num_no_good_cuts_added[
solve_data.LB]
for i in range(valid_no_good_cuts_num + 1,
len(MindtPy.MindtPy_linear_cuts.integer_cuts) + 1):
MindtPy.MindtPy_linear_cuts.integer_cuts[i].deactivate()
elif config.strategy == 'OA':
MindtPy.MindtPy_linear_cuts.integer_cuts[len(
MindtPy.MindtPy_linear_cuts.integer_cuts)].deactivate()
# MindtPy.MindtPy_linear_cuts.oa_cuts.activate()
masteropt = SolverFactory(config.mip_solver)
# determine if persistent solver is called.
if isinstance(masteropt, PersistentSolver):
masteropt.set_instance(solve_data.mip, symbolic_solver_labels=True)
mip_args = dict(config.mip_solver_args)
elapsed = get_main_elapsed_time(solve_data.timing)
remaining = int(max(config.time_limit - elapsed, 1))
if config.mip_solver == 'gams':
mip_args['add_options'] = mip_args.get('add_options', [])
mip_args['add_options'].append('option optcr=0.001;')
if config.threads > 0:
masteropt.options["threads"] = config.threads
master_mip_results = masteropt.solve(solve_data.mip,
tee=config.solver_tee,
**mip_args)
main_objective = next(
solve_data.working_model.component_data_objects(Objective,
active=True))
if main_objective.sense == minimize:
solve_data.LB = max([master_mip_results.problem.lower_bound] +
solve_data.LB_progress[:-1])
solve_data.LB_progress.append(solve_data.LB)
else:
solve_data.UB = min([master_mip_results.problem.upper_bound] +
solve_data.UB_progress[:-1])
solve_data.UB_progress.append(solve_data.UB)
config.logger.info('Fixed bound values: LB: {} UB: {}'.format(
solve_data.LB, solve_data.UB))
# Check bound convergence
if solve_data.LB + config.bound_tolerance >= solve_data.UB:
solve_data.results.solver.termination_condition = tc.optimal
def MindtPy_iteration_loop(solve_data, config):
"""
Main loop for MindtPy Algorithms
This is the outermost function for the algorithms in this package; this function controls the progression of
solving the model.
Parameters
----------
solve_data: MindtPy Data Container
data container that holds solve-instance data
config: ConfigBlock
contains the specific configurations for the algorithm
"""
working_model = solve_data.working_model
main_objective = next(
working_model.component_data_objects(Objective, active=True))
while solve_data.mip_iter < config.iteration_limit:
config.logger.info('---MindtPy Master Iteration %s---' %
solve_data.mip_iter)
solve_data.mip_subiter = 0
# solve MILP master problem
if config.strategy in {'OA', 'GOA', 'ECP'}:
master_mip, master_mip_results = solve_OA_master(
solve_data, config)
if config.single_tree is False:
if master_mip_results.solver.termination_condition is tc.optimal:
handle_master_mip_optimal(master_mip, solve_data, config)
elif master_mip_results.solver.termination_condition is tc.infeasible:
handle_master_mip_infeasible(master_mip, solve_data,
config)
last_iter_cuts = True
break
else:
handle_master_mip_other_conditions(master_mip,
master_mip_results,
solve_data, config)
# Call the MILP post-solve callback
config.call_after_master_solve(master_mip, solve_data)
else:
raise NotImplementedError()
if algorithm_should_terminate(solve_data, config, check_cycling=True):
last_iter_cuts = False
break
if config.single_tree is False and config.strategy != 'ECP': # if we don't use lazy callback, i.e. LP_NLP
# Solve NLP subproblem
# The constraint linearization happens in the handlers
fixed_nlp, fixed_nlp_result = solve_NLP_subproblem(
solve_data, config)
if fixed_nlp_result.solver.termination_condition in {
tc.optimal, tc.locallyOptimal, tc.feasible
}:
handle_NLP_subproblem_optimal(fixed_nlp, solve_data, config)
elif fixed_nlp_result.solver.termination_condition is tc.infeasible:
handle_NLP_subproblem_infeasible(fixed_nlp, solve_data, config)
else:
handle_NLP_subproblem_other_termination(
fixed_nlp, fixed_nlp_result.solver.termination_condition,
solve_data, config)
# Call the NLP post-solve callback
config.call_after_subproblem_solve(fixed_nlp, solve_data)
if algorithm_should_terminate(solve_data, config, check_cycling=False):
last_iter_cuts = True
break
if config.strategy == 'ECP':
add_ecp_cuts(solve_data.mip, solve_data, config)
# if config.strategy == 'PSC':
# # If the hybrid algorithm is not making progress, switch to OA.
# progress_required = 1E-6
# if main_objective.sense == minimize:
# log = solve_data.LB_progress
# sign_adjust = 1
# else:
# log = solve_data.UB_progress
# sign_adjust = -1
# # Maximum number of iterations in which the lower (optimistic)
# # bound does not improve before switching to OA
# max_nonimprove_iter = 5
# making_progress = True
# # TODO-romeo Unneccesary for OA and LOA, right?
# for i in range(1, max_nonimprove_iter + 1):
# try:
# if (sign_adjust * log[-i]
# <= (log[-i - 1] + progress_required)
# * sign_adjust):
# making_progress = False
# else:
# making_progress = True
# break
# except IndexError:
# # Not enough history yet, keep going.
# making_progress = True
# break
# if not making_progress and (
# config.strategy == 'hPSC' or
# config.strategy == 'PSC'):
# config.logger.info(
# 'Not making enough progress for {} iterations. '
# 'Switching to OA.'.format(max_nonimprove_iter))
# config.strategy = 'OA'
# if add_nogood_cuts is True, the bound obtained in the last iteration is no reliable.
# we correct it after the iteration.
if config.add_nogood_cuts:
bound_fix(solve_data, config, last_iter_cuts)
def MindtPy_initialize_master(solve_data, config):
"""
Initializes the decomposition algorithm and creates the master MIP/MILP problem.
This function initializes the decomposition problem, which includes generating the initial cuts required to
build the master MIP/MILP
Parameters
----------
solve_data: MindtPy Data Container
data container that holds solve-instance data
config: ConfigBlock
contains the specific configurations for the algorithm
"""
# if single tree is activated, we need to add bounds for unbounded variables in nonlinear constraints to avoid unbounded master problem.
if config.single_tree:
var_bound_add(solve_data, config)
m = solve_data.mip = solve_data.working_model.clone()
MindtPy = m.MindtPy_utils
if config.use_dual:
m.dual.deactivate()
if config.strategy == 'OA':
calc_jacobians(solve_data, config) # preload jacobians
MindtPy.MindtPy_linear_cuts.oa_cuts = ConstraintList(
doc='Outer approximation cuts')
elif config.strategy == 'ECP':
calc_jacobians(solve_data, config) # preload jacobians
MindtPy.MindtPy_linear_cuts.ecp_cuts = ConstraintList(
doc='Extended Cutting Planes')
# elif config.strategy == 'PSC':
# detect_nonlinear_vars(solve_data, config)
# MindtPy.MindtPy_linear_cuts.psc_cuts = ConstraintList(
# doc='Partial surrogate cuts')
# elif config.strategy == 'GBD':
# MindtPy.MindtPy_linear_cuts.gbd_cuts = ConstraintList(
# doc='Generalized Benders cuts')
# Set default initialization_strategy
if config.init_strategy is None:
if config.strategy in {'OA', 'GOA'}:
config.init_strategy = 'rNLP'
else:
config.init_strategy = 'max_binary'
config.logger.info(
'{} is the initial strategy being used.'
'\n'.format(
config.init_strategy))
# Do the initialization
if config.init_strategy == 'rNLP':
init_rNLP(solve_data, config)
elif config.init_strategy == 'max_binary':
init_max_binaries(solve_data, config)
elif config.init_strategy == 'initial_binary':
if config.strategy != 'ECP':
fixed_nlp, fixed_nlp_result = solve_NLP_subproblem(
solve_data, config)
if fixed_nlp_result.solver.termination_condition in {tc.optimal, tc.locallyOptimal, tc.feasible}:
handle_NLP_subproblem_optimal(fixed_nlp, solve_data, config)
elif fixed_nlp_result.solver.termination_condition is tc.infeasible:
handle_NLP_subproblem_infeasible(fixed_nlp, solve_data, config)
else:
handle_NLP_subproblem_other_termination(fixed_nlp, fixed_nlp_result.solver.termination_condition,
solve_data, config)