def __init__(self, model):
self.model = model
self.original_to_new_var_map = pe.ComponentMap()
self.aux_var_map = dict()
self.reverse_var_map = dict()
self.var_relax_map = pe.ComponentMap()
self.degree_map = pe.ComponentMap()
self.counter = RelaxationCounter()
def set_input(self,
aux_var,
shape,
f_x_expr,
persistent_solvers=None,
large_eval_tol=math.inf,
use_linear_relaxation=True):
"""
Parameters
----------
aux_var: pyomo.core.base.var._GeneralVarData
The auxiliary variable replacing f(x)
shape: FunctionShape
Either FunctionShape.CONVEX or FunctionShape.CONCAVE
f_x_expr: pyomo expression
The pyomo expression representing f(x)
persistent_solvers: list
List of persistent solvers that should be updated when the relaxation changes
large_eval_tol: float
To avoid numerical problems, if f_x_expr or its derivative evaluates to a value larger than large_eval_tol,
at a point in x_pts, then that point is skipped.
use_linear_relaxation: bool
Specifies whether a linear or nonlinear relaxation should be used
"""
if shape not in {FunctionShape.CONVEX, FunctionShape.CONCAVE}:
raise ValueError(
'MultivariateRelaxation only supports concave or convex functions.'
)
self._function_shape = shape
if shape == FunctionShape.CONVEX:
relaxation_side = RelaxationSide.UNDER
else:
relaxation_side = RelaxationSide.OVER
self._set_input(relaxation_side=relaxation_side,
persistent_solvers=persistent_solvers,
use_linear_relaxation=use_linear_relaxation,
large_eval_tol=large_eval_tol)
self._xs = list(identify_variables(f_x_expr, include_fixed=False))
self._aux_var_ref.set_component(aux_var)
self._f_x_expr = f_x_expr
lb_oa_pt = pe.ComponentMap()
ub_oa_pt = pe.ComponentMap()
should_use_lb_oa_pt = True
should_use_ub_oa_pt = True
for v in self._xs:
lb, ub = tuple(_get_bnds_list(v))
if lb <= -math.inf:
should_use_lb_oa_pt = False
else:
lb_oa_pt[v] = lb
if ub >= math.inf:
should_use_ub_oa_pt = False
else:
ub_oa_pt[v] = ub
if should_use_lb_oa_pt:
self.add_oa_point(var_values=lb_oa_pt)
if should_use_ub_oa_pt:
self.add_oa_point(var_values=ub_oa_pt)
def test_push_and_pop_partitions_2(self):
m = pe.ConcreteModel()
m.x = pe.Var(bounds=(-1, 1))
m.y = pe.Var()
m.c = coramin.relaxations.PWXSquaredRelaxation()
m.c.build(x=m.x, aux_var=m.y)
self.assertEqual(m.c._partitions, pe.ComponentMap([(m.x, [-1, 1])]))
m.x.setlb(0)
m.c.rebuild()
self.assertEqual(m.c._partitions, pe.ComponentMap([(m.x, [0, 1])]))
m.x.setlb(-1)
m.c.rebuild()
self.assertEqual(m.c._partitions, pe.ComponentMap([(m.x, [-1, 1])]))
m.x.value = 0.5
m.c.add_partition_point()
m.c.rebuild()
self.assertEqual(m.c._partitions, pe.ComponentMap([(m.x, [-1, 0.5, 1])]))
m.x.setlb(0)
m.c.rebuild()
self.assertEqual(m.c._partitions, pe.ComponentMap([(m.x, [0, 0.5, 1])]))
m.x.setlb(-1)
m.c.rebuild()
self.assertEqual(m.c._partitions, pe.ComponentMap([(m.x, [-1, 0.5, 1])]))
m.x.setub(0)
m.c.rebuild()
self.assertEqual(m.c._partitions, pe.ComponentMap([(m.x, [-1, 0])]))
m.x.setub(1)
m.c.rebuild()
self.assertEqual(m.c._partitions, pe.ComponentMap([(m.x, [-1, 1])]))
def relax_expression(model, expr, relaxation_side, data):
relaxation_side_map = pe.ComponentMap()
relaxation_side_map[expr] = relaxation_side
expr = _relax_expr(expr=expr, aux_var_map=data.aux_var_map, parent_block=model,
relaxation_side_map=relaxation_side_map, counter=data.counter, degree_map=data.degree_map)
return expr
def _solution_from_tree(self, problem, tree):
nodes_visited = tree.nodes_visited
if len(tree.solution_pool) == 0:
# Return lower bound only
optimal_vars = pe.ComponentMap(
(v, pe.value(v))
for v in problem.component_data_objects(pe.Var, active=True))
return BabSolution(
BabStatusInterrupted(),
None,
optimal_vars,
dual_bound=tree.state.lower_bound,
nodes_visited=nodes_visited,
)
primal_solution = tree.solution_pool.head
return BabSolution(
primal_solution.status,
primal_solution.objective,
primal_solution.variables,
dual_bound=tree.state.lower_bound,
nodes_visited=nodes_visited,
nodes_remaining=len(tree.open_nodes),
is_timeout=self.galini.timelimit.timeout(),
has_converged=self.has_converged(tree.state),
node_limit_exceeded=self.node_limit_exceeded(tree.state),
)
def add_oa_point(self, var_values=None):
"""
Add a point at which an outer-approximation cut for a convex constraint should be added. This does not
rebuild the relaxation. You must call rebuild() for the constraint to get added.
Parameters
----------
var_values: pe.ComponentMap
"""
if var_values is None:
var_values = pe.ComponentMap()
for v in self.get_rhs_vars():
var_values[v] = v.value
else:
var_values = pe.ComponentMap(var_values)
self._oa_points.append(var_values)
def test_adjacency_matrix(galini, problem):
linear_model, _, _ = relax(problem)
galini.timelimit.start_now()
triangle_cuts_gen = TriangleCutsGenerator(
galini, galini._config.cuts_generator.triangle)
triangle_cuts_gen.before_start_at_root(problem, linear_model)
lower_bounds, upper_bounds, domains, aux_vars, var_by_id, edges = \
triangle_cuts_gen._detect_bilinear_terms(linear_model)
expected_adj = [[1, 1, 1, 0, 0, 1, 0, 1], [1, 0, 1, 1, 1, 0, 1, 1],
[1, 1, 1, 1, 0, 1, 1, 1], [0, 1, 1, 1, 1, 1, 1, 1],
[0, 1, 0, 1, 0, 1, 1, 1], [1, 0, 1, 1, 1, 1, 1, 1],
[0, 1, 1, 1, 1, 1, 0, 0], [1, 1, 1, 1, 1, 1, 0, 1]]
var_to_idx = aml.ComponentMap()
for i in problem.I:
x = linear_model.x[i]
var_to_idx[x] = i
for (x_id, y_id) in edges:
x = var_by_id[x_id]
y = var_by_id[y_id]
x_idx = var_to_idx[x]
y_idx = var_to_idx[y]
assert expected_adj[x_idx][y_idx] == 1
# Mark edge as visited
expected_adj[x_idx][y_idx] = 0
# Check we visited all edges
assert np.all(np.isclose(expected_adj, 0))
def set_input(self, master_vars, tol=1e-6, comm = None):
"""
It is very important for master_vars to be in the same order for every process.
Parameters
----------
master_vars
tol
"""
self.comm = None
if comm is not None:
self.comm = comm
else:
self.comm = MPI.COMM_WORLD
self.num_subproblems_by_rank = np.zeros(self.comm.Get_size())
del self.cuts
self.cuts = pe.ConstraintList()
self.subproblems = list()
self.master_etas = list()
self.complicating_vars_maps = list()
self.master_vars = list(master_vars)
self.master_vars_indices = pe.ComponentMap()
for i, v in enumerate(self.master_vars):
self.master_vars_indices[v] = i
self.tol = tol
self.subproblem_solvers = list()
self.all_master_etas = list()
self._subproblem_ndx_map = dict()
def solve(self, model, linear_model, mip_solution, tree, node):
assert mip_solution.status.is_success(), "Should be a feasible point for the relaxation"
model_var_map = node.storage.model_to_relaxation_var_map
mip_solution_with_model_vars = pe.ComponentMap(
(var, mip_solution.variables[model_var_map[var]])
for var in model.component_data_objects(pe.Var, active=True)
)
self.algorithm._update_solver_options(self.algorithm._nlp_solver)
primal_solution = solve_primal_with_starting_point(
model, mip_solution_with_model_vars, self.algorithm._nlp_solver, self.galini.mc, fix_all=True
)
if primal_solution is not None and primal_solution.status.is_success():
return primal_solution
self.algorithm._update_solver_options(self.algorithm._nlp_solver)
new_primal_solution = solve_primal(
model, mip_solution_with_model_vars, self.algorithm._nlp_solver, self.galini.mc
)
if new_primal_solution is not None:
primal_solution = new_primal_solution
return primal_solution
def push_oa_points(self, key=None):
"""
Save the current list of OA points for later use through pop_oa_points().
"""
to_save = [pe.ComponentMap(i) for i in self._oa_points]
if key is not None:
self._oa_stack_map[key] = to_save
else:
self._saved_oa_points.append(to_save)
def attach_nonant_var_map(self, scenario_name):
instance = self.local_scenarios[scenario_name]
subproblem_to_root_vars_map = pyo.ComponentMap()
for var, rvar in zip(instance._mpisppy_data.nonant_indices.values(), self.root_vars):
if var.name not in rvar.name:
raise Exception("Error: Complicating variable mismatch, sub-problem variables changed order")
subproblem_to_root_vars_map[var] = rvar
# this is for interefacing with PH code
instance._mpisppy_model.subproblem_to_root_vars_map = subproblem_to_root_vars_map
def optimization_data(self, model):
data = pyo.ComponentMap()
fs = model.fs
data[fs.product.flow_mass_phase_comp[0, "Liq", "H2O"]] = 0.385923
data[fs.product.flow_mass_phase_comp[0, "Liq", "NaCl"]] = 0.4824263e-3
data[fs.disposal.flow_mass_phase_comp[0, "Liq", "H2O"]] = 0.544077
data[fs.disposal.flow_mass_phase_comp[0, "Liq", "NaCl"]] = 0.695177e-1
data[fs.costing.LCOW] = 1.13407
data[fs.water_recovery] = 0.386405
return data
def simulation_data(self, model):
data = pyo.ComponentMap()
fs = model.fs
data[fs.product.flow_mass_phase_comp[0, 'Liq', 'H2O']] = 0.179331
data[fs.product.flow_mass_phase_comp[0, 'Liq', 'NaCl']] = 0.286037e-3
data[fs.disposal.flow_mass_phase_comp[0, 'Liq', 'H2O']] = 0.750668
data[fs.disposal.flow_mass_phase_comp[0, 'Liq', 'NaCl']] = 0.697139e-1
data[fs.costing.LCOW] = 1.73465
data[fs.water_recovery] = 0.179618
return data
def optimization_data(self, model):
data = pyo.ComponentMap()
fs = model.fs
data[fs.product.flow_mass_phase_comp[0, "Liq", "H2O"]] = 0.732036
data[fs.product.flow_mass_phase_comp[0, "Liq", "NaCl"]] = 0.446025e-3
data[fs.disposal.flow_mass_phase_comp[0, "Liq", "H2O"]] = 0.197967
data[fs.disposal.flow_mass_phase_comp[0, "Liq", "NaCl"]] = 0.695539e-1
data[fs.costing.LCOW] = 1.48679
data[fs.water_recovery] = 0.732481
return data
def initialization_data(self, model):
data = pyo.ComponentMap()
fs = model.fs
data[fs.product.flow_mass_phase_comp[0, "Liq", "H2O"]] = 0.351684
data[fs.product.flow_mass_phase_comp[0, "Liq", "NaCl"]] = 0.269335e-3
data[fs.disposal.flow_mass_phase_comp[0, "Liq", "H2O"]] = 0.606601
data[fs.disposal.flow_mass_phase_comp[0, "Liq", "NaCl"]] = 0.695767e-1
data[fs.costing.LCOW] = 1.66166
data[fs.water_recovery] = 0.5
return data
def optimization_data(self, model):
data = pyo.ComponentMap()
fs = model.fs
data[fs.product.flow_mass_phase_comp[0, "Liq", "H2O"]] = 0.732053
data[fs.product.flow_mass_phase_comp[0, "Liq", "NaCl"]] = 0.454317e-3
data[fs.disposal.flow_mass_phase_comp[0, "Liq", "H2O"]] = 0.197952
data[fs.disposal.flow_mass_phase_comp[0, "Liq", "NaCl"]] = 0.695457e-1
data[fs.costing.LCOW] = 1.18849
data[fs.water_recovery] = 0.732504
return data
def simulation_data(self, model):
data = pyo.ComponentMap()
fs = model.fs
data[fs.product.flow_mass_phase_comp[0, "Liq", "H2O"]] = 0.296269
data[fs.product.flow_mass_phase_comp[0, "Liq", "NaCl"]] = 0.274578e-3
data[fs.disposal.flow_mass_phase_comp[0, "Liq", "H2O"]] = 0.633730
data[fs.disposal.flow_mass_phase_comp[0, "Liq", "NaCl"]] = 0.697254e-1
data[fs.costing.LCOW] = 1.73385
data[fs.water_recovery] = 0.296544
return data
def initialization_data(self, model):
data = pyo.ComponentMap()
fs = model.fs
data[fs.product.flow_mass_phase_comp[0, "Liq", "H2O"]] = 0.298029
data[fs.product.flow_mass_phase_comp[0, "Liq", "NaCl"]] = 0.274413e-3
data[fs.disposal.flow_mass_phase_comp[0, "Liq", "H2O"]] = 0.633927
data[fs.disposal.flow_mass_phase_comp[0, "Liq", "NaCl"]] = 0.697161e-1
data[fs.costing.LCOW] = 1.66663
data[fs.water_recovery] = 0.5
return data
def optimization_data(self, model):
data = pyo.ComponentMap()
fs = model.fs
data[fs.product.flow_mass_phase_comp[0, 'Liq', 'H2O']] = 0.732053
data[fs.product.flow_mass_phase_comp[0, 'Liq', 'NaCl']] = 0.456208e-3
data[fs.disposal.flow_mass_phase_comp[0, 'Liq', 'H2O']] = 0.197952
data[fs.disposal.flow_mass_phase_comp[0, 'Liq', 'NaCl']] = 0.695438e-1
data[fs.costing.LCOW] = 1.17018
data[fs.water_recovery] = 0.732504
return data
def simulation_data(self, model):
data = pyo.ComponentMap()
fs = model.fs
data[fs.product.flow_mass_phase_comp[0, "Liq", "H2O"]] = 0.179331
data[fs.product.flow_mass_phase_comp[0, "Liq", "NaCl"]] = 0.286037e-3
data[fs.disposal.flow_mass_phase_comp[0, "Liq", "H2O"]] = 0.750668
data[fs.disposal.flow_mass_phase_comp[0, "Liq", "NaCl"]] = 0.697139e-1
data[fs.costing.LCOW] = 2.00482
data[fs.water_recovery] = 0.179618
return data
def optimization_data(self, model):
data = pyo.ComponentMap()
fs = model.fs
data[fs.product.flow_mass_phase_comp[0, 'Liq', 'H2O']] = 0.732036
data[fs.product.flow_mass_phase_comp[0, 'Liq', 'NaCl']] = 0.448848e-3
data[fs.disposal.flow_mass_phase_comp[0, 'Liq', 'H2O']] = 0.197967
data[fs.disposal.flow_mass_phase_comp[0, 'Liq', 'NaCl']] = 0.695512e-1
data[fs.costing.LCOW] = 1.46933
data[fs.water_recovery] = 0.732481
return data
def simulation_data(self, model):
data = pyo.ComponentMap()
fs = model.fs
data[fs.product.flow_mass_phase_comp[0, "Liq", "H2O"]] = 0.329390
data[fs.product.flow_mass_phase_comp[0, "Liq", "NaCl"]] = 0.271454e-3
data[fs.disposal.flow_mass_phase_comp[0, "Liq", "H2O"]] = 0.600609
data[fs.disposal.flow_mass_phase_comp[0, "Liq", "NaCl"]] = 0.697285e-1
data[fs.costing.LCOW] = 1.86314
data[fs.water_recovery] = 0.329661
return data
def create_subproblem(root):
m = pyo.ConcreteModel()
m.x1 = pyo.Var()
m.x2 = pyo.Var()
m.y = pyo.Var()
m.obj = pyo.Objective(expr=-m.x2)
m.c1 = pyo.Constraint(expr=(m.x1 - 1)**2 + m.x2**2 <= pyo.log(m.y))
m.c2 = pyo.Constraint(expr=(m.x1 + 1)**2 + m.x2**2 <= pyo.log(m.y))
complicating_vars_map = pyo.ComponentMap()
complicating_vars_map[root.y] = m.y
return m, complicating_vars_map
def create_subproblem(root, farmer, scenario):
m = pyo.ConcreteModel()
m.crops = pyo.Set(initialize=farmer.crops, ordered=True)
m.devoted_acreage = pyo.Var(m.crops)
m.QuantitySubQuotaSold = pyo.Var(m.crops, bounds=(0.0, None))
m.QuantitySuperQuotaSold = pyo.Var(m.crops, bounds=(0.0, None))
m.QuantityPurchased = pyo.Var(m.crops, bounds=(0.0, None))
def EnforceCattleFeedRequirement_rule(m, i):
return (
farmer.CattleFeedRequirement[i] <=
(farmer.crop_yield[scenario][i] * m.devoted_acreage[i]) +
m.QuantityPurchased[i] - m.QuantitySubQuotaSold[i] -
m.QuantitySuperQuotaSold[i])
m.EnforceCattleFeedRequirement = pyo.Constraint(
m.crops, rule=EnforceCattleFeedRequirement_rule)
def LimitAmountSold_rule(m, i):
return m.QuantitySubQuotaSold[i] + m.QuantitySuperQuotaSold[
i] - (farmer.crop_yield[scenario][i] *
m.devoted_acreage[i]) <= 0.0
m.LimitAmountSold = pyo.Constraint(m.crops,
rule=LimitAmountSold_rule)
def EnforceQuotas_rule(m, i):
return (0.0, m.QuantitySubQuotaSold[i], farmer.PriceQuota[i])
m.EnforceQuotas = pyo.Constraint(m.crops, rule=EnforceQuotas_rule)
obj_expr = sum(farmer.PurchasePrice[crop] *
m.QuantityPurchased[crop] for crop in m.crops)
obj_expr -= sum(farmer.SubQuotaSellingPrice[crop] *
m.QuantitySubQuotaSold[crop] for crop in m.crops)
obj_expr -= sum(farmer.SuperQuotaSellingPrice[crop] *
m.QuantitySuperQuotaSold[crop] for crop in m.crops)
m.obj = pyo.Objective(
expr=farmer.scenario_probabilities[scenario] * obj_expr)
complicating_vars_map = pyo.ComponentMap()
for crop in m.crops:
complicating_vars_map[
root.devoted_acreage[crop]] = m.devoted_acreage[crop]
return m, complicating_vars_map
def branch_at_point(model, current_bounds, branching_point, mc):
"""Branch problem at branching_point, returning a list of child problems."""
current_bounds = pe.ComponentMap()
for var in model.component_data_objects(pe.Var,
active=True,
descend_into=True):
var_lb = var.lb
var_ub = var.ub
if var_lb is None:
var_lb = -np.inf
if var_ub is None:
var_ub = np.inf
current_bounds[var] = (var_lb, var_ub)
var = branching_point.variable
var_lb, var_ub = current_bounds[var]
epsilon = mc.epsilon
for point in branching_point.points:
is_less_than_ub = almost_le(point, var_ub, atol=epsilon)
is_greater_than_lb = almost_ge(point, var_lb, atol=epsilon)
if not is_less_than_ub or not is_greater_than_lb:
raise RuntimeError(
'Branching outside variable bounds: {} in [{}, {}], branching at {}'
.format(var.name, var.lb, var.ub, point))
children = []
new_upper_bound = var_lb
is_integer = var.is_integer() or var.is_binary()
for point in branching_point.points:
new_lower_bound = new_upper_bound
new_upper_bound = point
var_lower_bound = \
np.ceil(new_lower_bound) if is_integer else new_lower_bound
var_upper_bound = \
np.floor(new_upper_bound) if is_integer else new_upper_bound
child = copy.copy(current_bounds)
child[var] = (var_lower_bound, var_upper_bound)
children.append(child)
var_lower_bound = \
np.ceil(new_upper_bound) if is_integer else new_upper_bound
var_upper_bound = \
np.floor(var_ub) if is_integer else var_ub
child = copy.copy(current_bounds)
child[var] = (var_lower_bound, var_upper_bound)
children.append(child)
return children
def _cache_and_set_relaxed_bounds(self, bound_relax_factor):
self._bound_cache = pyo.ComponentMap()
val = pyo.value
for v in self._model.component_data_objects(pyo.Var, active=True, descend_into=True):
# we could hit a variable more
# than once because of References
if v in self._bound_cache:
continue
if v.lb is None and v.ub is None:
continue
self._bound_cache[v] = (v.lb, v.ub)
sf = get_scaling_factor(v, default=1)
if v.lb is not None:
v.lb = val((v.lb*sf - bound_relax_factor*max(1, abs(val(v.lb*sf))))/sf)
if v.ub is not None:
v.ub = val((v.ub*sf + bound_relax_factor*max(1, abs(val(v.ub*sf))))/sf)
def _check_if_problem_is_nolinear(self, relaxed_problem):
self._convex_relaxations_map = pe.ComponentMap()
for relaxation in relaxed_problem.galini_nonlinear_relaxations:
is_convex = (relaxation.is_rhs_convex()
and relaxation.relaxation_side in [
RelaxationSide.BOTH, RelaxationSide.UNDER
]) or (relaxation.is_rhs_concave()
and relaxation.relaxation_side
in [RelaxationSide.BOTH, RelaxationSide.OVER])
if is_convex:
rhs_expr = relaxation.get_rhs_expr()
rhs_vars = relaxation.get_rhs_vars()
aux_var = relaxation.get_aux_var()
rel_expr = rhs_expr - aux_var
rel_vars = rhs_vars + [aux_var]
self._convex_relaxations_map[relaxation] = (rel_expr, rel_vars)
def __init__(self, component):
if not mpi4py_available:
raise ImportError('BendersCutGenerator requires mpi4py.')
if not numpy_available:
raise ImportError('BendersCutGenerator requires numpy.')
_BlockData.__init__(self, component)
self.num_subproblems_by_rank = np.zeros(MPI.COMM_WORLD.Get_size())
self.subproblems = list()
self.complicating_vars_maps = list()
self.master_vars = list()
self.master_vars_indices = pe.ComponentMap()
self.master_etas = list()
self.cuts = None
self.subproblem_solvers = list()
self.tol = None
self.all_master_etas = list()
def solution_pool_from_solver(solver):
if not isinstance(solver, CPLEXDirect):
return None
solver_model = solver._solver_model
solver_pool = solver_model.solution.pool
num_sol = solver_pool.get_num()
if not num_sol:
return None
pool = []
var_map = solver._pyomo_var_to_ndx_map
for i in range(num_sol):
values = solver_pool.get_values(i)
vars_to_load = var_map.keys()
sol = pe.ComponentMap(
(var, value) for var, value in zip(vars_to_load, values))
pool.append(sol)
return pool
def __init__(self, component):
if not mpi4py_available:
raise ImportError('BendersCutGenerator requires mpi4py.')
if not numpy_available:
raise ImportError('BendersCutGenerator requires numpy.')
_BlockData.__init__(self, component)
self.num_subproblems_by_rank = 0 #np.zeros(self.comm.Get_size())
self.subproblems = list()
self.complicating_vars_maps = list()
self.master_vars = list()
self.master_vars_indices = pe.ComponentMap()
self.master_etas = list()
self.cuts = None
self.subproblem_solvers = list()
self.tol = None
self.all_master_etas = list()
self._subproblem_ndx_map = dict() # map from ndx in self.subproblems (local) to the global subproblem ndx