Source code for gemseo.core.derivatives.mda_derivatives
# Copyright 2021 IRT Saint Exupéry, https://www.irt-saintexupery.com
#
# This program is free software; you can redistribute it and/or
# modify it under the terms of the GNU Lesser General Public
# License version 3 as published by the Free Software Foundation.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
# Lesser General Public License for more details.
#
# You should have received a copy of the GNU Lesser General Public License
# along with this program; if not, write to the Free Software Foundation,
# Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
# Contributors:
# INITIAL AUTHORS - API and implementation and/or documentation
# :author: Francois Gallard
# OTHER AUTHORS - MACROSCOPIC CHANGES
"""Graph algorithms to solve the Jacobian accumulation problem for MDOChain and MDA."""
from __future__ import annotations
import uuid
from typing import TYPE_CHECKING
from gemseo.core.coupling_structure import MDOCouplingStructure
from gemseo.core.derivatives.chain_rule import traverse_add_diff_io
from gemseo.core.discipline import MDODiscipline
if TYPE_CHECKING:
from collections.abc import Iterable
from gemseo.core.derivatives.chain_rule import DisciplineIOMapping
def _replace_strongly_coupled(
coupling_structure: MDOCouplingStructure,
) -> tuple[list[MDODiscipline], list[MDODiscipline]]:
"""Replace the strongly coupled disciplines by a single one.
The replacing discipline has for inputs the merged inputs of all coupled
discipline except for the strong couplings, while its outputs are all combined
outputs.
Args:
coupling_structure: The input coupling structure containing the graph and all
the disciplines
Returns:
All the disciplines with the strongly coupled replaced by the
merged one, and the replacing disciplines.
"""
disciplines_with_group = list(coupling_structure.disciplines)
reduced_disciplines = []
all_disc_with_red = []
strong_c = set(coupling_structure.strong_couplings)
for parallel_tasks in coupling_structure.sequence:
for group in parallel_tasks:
# The strong coupling cycles are treated here
# And also the self coupled disciplines
if len(group) > 1 or (
len(group) == 1 and coupling_structure.is_self_coupled(group[0])
):
disc_merged = MDODiscipline(str(uuid.uuid4()))
for disc in group:
disciplines_with_group.remove(disc)
# The strong couplings are not real dependencies of the MDA for
# derivatives computation.
disc_merged.input_grammar.update_from_names(
set(disc.input_grammar.names) - strong_c
)
disc_merged.output_grammar.update_from_names(
disc.output_grammar.names
)
all_disc_with_red.append(disc_merged)
reduced_disciplines.append(disc_merged)
else:
disc = group[0]
# The weakly coupled disciplines are treated like in the chain.
all_disc_with_red.append(disc)
return all_disc_with_red, reduced_disciplines
[docs]
def traverse_add_diff_io_mda(
coupling_structure: MDOCouplingStructure,
inputs: Iterable[str],
outputs: Iterable[str],
) -> DisciplineIOMapping:
"""Set the required differentiated IOs for the disciplines in a chain.
Add the differentiated inputs and outputs to the disciplines in a chain of of
executions, given the list of inputs with respect to which the derivation is made
and the list of outputs to be derived. The inputs and outputs are a subset of all
the inputs and outputs of the chain. This allows to minimize the number of
linearizations to perform in the disciplines.
Uses a two ways (from inputs to outputs and then from outputs to inputs)
breadth first search graph traversal algorithm.
The graph is constructed by :class:`.DependencyGraph`, which represents the data
connexions (edges) between the disciplines (nodes).
Args:
coupling_structure: The coupling structure of the MDA.
inputs: The inputs with respect to which the chain chain is differentiated.
outputs: The chain outputs to be differentiated.
Returns:
The merged differentiated inputs and outputs.
"""
strong_groups = coupling_structure.get_strongly_coupled_disciplines(
by_group=True, add_self_coupled=True
)
all_disc_with_red, reduced_disciplines = _replace_strongly_coupled(
coupling_structure
)
reduced_coupling_structure = MDOCouplingStructure(all_disc_with_red)
diff_ios_merged = traverse_add_diff_io(
reduced_coupling_structure.graph.graph,
inputs,
outputs,
add_differentiated_ios=True,
)
# The sub MDAs where the strong couplings are handled here.
strong_couplings = coupling_structure.strong_couplings
for group, disc_reduced in zip(strong_groups, reduced_disciplines):
if disc_reduced in diff_ios_merged:
diff_red_in = set(diff_ios_merged[disc_reduced][0])
diff_red_out = set(diff_ios_merged[disc_reduced][1])
for disc in group:
# There is a need to differentiate with respect to all the inputs of
# the MDA that are also inputs of the discipline
# And we add all strong input couplings
# Finally we keep only the discipline inputs.
diff_in = diff_red_in.union(strong_couplings).intersection(
disc.input_grammar.names
)
disc.add_differentiated_inputs(diff_in)
# It is simpler for the outputs because the outputs to be differentiated
# are the ones from the MDA.
diff_out = diff_red_out.union(strong_couplings).intersection(
disc.output_grammar.names
)
disc.add_differentiated_outputs(diff_out)
diff_ios_merged[disc] = (list(diff_in), list(diff_out))
# The state variables and residuals must be differentiated too, only for the
# disciplines involved in the computations.
for disc, diffio_disc in diff_ios_merged.items():
if disc.residual_variables:
residuals = disc.residual_variables.keys()
states = disc.residual_variables.values()
diffio_disc[0].extend(states)
diffio_disc[1].extend(residuals)
disc.add_differentiated_inputs(states)
disc.add_differentiated_outputs(residuals)
return diff_ios_merged
