Add RIA, PMA, SORA based on https://github.com/l-brevault/RBDO_OT#48
Open
regislebrun wants to merge 1 commit into
Open
Add RIA, PMA, SORA based on https://github.com/l-brevault/RBDO_OT#48regislebrun wants to merge 1 commit into
regislebrun wants to merge 1 commit into
Conversation
|
Warning Review limit reached
Next review available in: 52 minutes Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available. How can I continue?After more reviews become available, a review can be triggered using the To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews. How do review limits work?CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability. For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window. Please refer docs for additional details. Review details⚙️ Run configurationConfiguration used: defaults Review profile: CHILL Plan: Pro Plus Run ID: 📒 Files selected for processing (47)
📝 WalkthroughWalkthroughAdds ChangesRBDO Problem and Algorithms
Estimated code review effort: 4 (Complex) | ~60 minutes Sequence Diagram(s)sequenceDiagram
participant Client
participant RBDOProblem
participant RIAAlgorithm
participant PMAAlgorithm
participant SORAAlgorithm
participant OptimizationSolver
Client->>RBDOProblem: configure objective, constraint, distributions, beta
Client->>RIAAlgorithm: run()
RIAAlgorithm->>OptimizationSolver: solve with FORM reliability constraint
OptimizationSolver-->>RIAAlgorithm: OptimizationResult
Client->>PMAAlgorithm: run()
PMAAlgorithm->>OptimizationSolver: solve with inverse-FORM constraint
OptimizationSolver-->>PMAAlgorithm: OptimizationResult
Client->>SORAAlgorithm: run()
SORAAlgorithm->>OptimizationSolver: solve deterministic shifted problem
SORAAlgorithm->>SORAAlgorithm: update most probable target points
OptimizationSolver-->>SORAAlgorithm: OptimizationResult
🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 11
♻️ Duplicate comments (1)
lib/src/SORAAlgorithm.cxx (1)
346-356: 🩺 Stability & Availability | 🟠 Major | ⚡ Quick winNo error handling for inverse FORM convergence failures.
Same issue as in
PMAAlgorithm.cxx(lines 258-268): the NLopt inverse FORM result is used without checking convergence. If the inner optimization fails, the MPTP and shifting vector will be garbage, causing the SORA iteration to diverge silently.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@lib/src/SORAAlgorithm.cxx` around lines 346 - 356, The inverse FORM solve in SORAAlgorithm::... uses NLopt::run() without checking whether the optimization converged. After invAlgo.run(), inspect the result/status from the inverse problem (as done in PMAAlgorithm::...) and only use the MPTP and shifting vector when convergence succeeds; otherwise stop the SORA iteration or propagate an explicit failure so bad values do not silently drive divergence.
🧹 Nitpick comments (3)
lib/src/SORAAlgorithm.cxx (1)
260-344: 📐 Maintainability & Code Quality | 🔵 Trivial | 🏗️ Heavy liftSignificant code duplication between PMA and SORA inverse FORM logic.
The
PMAObjective(lines 260-302) andNormConstraint(lines 309-344) classes are nearly identical toInverseFormObjective(PMAAlgorithm.cxx lines 170-213) andSphereConstraint(PMAAlgorithm.cxx lines 220-255). The inverse FORM setup code (NLopt, bounds, starting point, iteration limit) is also duplicated.Consider extracting a shared inverse FORM solver utility (e.g.,
solveInverseFORM(gParam, inverseTransform, op, targetBeta, dimXZ)) that both algorithms can call. This would reduce maintenance burden and ensure consistent error handling.♻️ Proposed extraction
// In a shared header, e.g., InverseFORMSolver.hxx struct InverseFORMResult { Point uMPTP; Point xzMPTP; Scalar gMPTP; bool converged; }; InverseFORMResult solveInverseFORM( const Function & gParam, const Function & inverseTransform, const ComparisonOperator & op, Scalar targetBeta, UnsignedInteger dimXZ);Both
PMAAlgorithm::run()andSORAAlgorithm::run()would call this utility instead of defining their own local classes.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@lib/src/SORAAlgorithm.cxx` around lines 260 - 344, Refactor the duplicated inverse FORM setup in SORAAlgorithm by extracting a shared solver utility that both SORA and PMA can reuse. The local helper classes PMAObjective and NormConstraint should be moved into a common implementation alongside the equivalent InverseFormObjective and SphereConstraint logic, and the repeated NLopt configuration, bounds, starting point, and iteration-limit code should be centralized into a reusable solveInverseFORM helper. Ensure SORAAlgorithm::run() calls the shared utility instead of defining its own near-identical objective/constraint classes.python/src/SORAAlgorithm_doc.i (1)
1-17: 📐 Maintainability & Code Quality | 🔵 Trivial | ⚡ Quick winDocument the
RBDOProblemconstructor in the class docstring.The class-level docstring only documents the
RobustOptimizationProblemconstructor, but the C++ header declares a second constructor takingRBDOProblem. Both test files constructSORAAlgorithmexclusively withRBDOProblem, making it the primary use case. Users relying on the docstring won't know this constructor exists.📝 Suggested docstring update
Parameters ---------- -problem : :class:`~otrobopt.RobustOptimizationProblem` - Robust optimization problem +problem : :class:`~otrobopt.RBDOProblem` or :class:`~otrobopt.RobustOptimizationProblem` + RBDO problem or robust optimization problem solver : :py:class:`openturns.OptimizationAlgorithm` Optimization solver🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@python/src/SORAAlgorithm_doc.i` around lines 1 - 17, The SORAAlgorithm class docstring only describes the RobustOptimizationProblem constructor, but it also supports an RBDOProblem constructor. Update the %feature("docstring") OTROBOPT::SORAAlgorithm text to document the RBDOProblem initialization path alongside the existing one, using the class name and constructor parameter names so users can see both supported ways to instantiate SORAAlgorithm.lib/test/t_RIAAlgorithm_std.cxx (1)
84-92: 🎯 Functional Correctness | 🔵 Trivial | ⚡ Quick winAdd assertions to validate the optimization result.
The comment at lines 87-88 states the expected result (
d1 + d2 ≈ 2*beta/sqrt(2) ≈ 4.24), but no assertion checks this. The test only prints the value, so it will pass even if the algorithm returns a grossly incorrect result. Adding a simple tolerance-based check would catch regressions without being overly brittle.💡 Suggested assertion
if (xStar.getDimension() == 2) { - std::cout << "d1+d2=" << xStar[0] + xStar[1] << std::endl; + const Scalar sum = xStar[0] + xStar[1]; + std::cout << "d1+d2=" << sum << std::endl; + if (std::abs(sum - 4.24) > 1.0) + { + throw Exception(HERE) << "RIA result d1+d2=" << sum << " is too far from expected ~4.24"; + } }🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@lib/test/t_RIAAlgorithm_std.cxx` around lines 84 - 92, The RIA test currently only prints the computed optimum and does not verify it, so add a tolerance-based assertion in the xStar check path to validate the expected d1+d2 value against the commented target. Use the existing ria.getResult().getOptimalPoint() and xStar.getDimension() logic to locate the spot, and assert that xStar[0] + xStar[1] is close to the expected 2*beta/sqrt(2) value with a reasonable tolerance so the test fails on grossly incorrect results without being brittle.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@lib/src/PMAAlgorithm.cxx`:
- Around line 257-281: The inverse FORM solve in PMAAlgorithm::run uses NLopt
results immediately after algo.run() without checking whether the optimizer
actually converged, so add explicit convergence/error handling around the NLopt
optimization before reading getResult(). Verify the solver status from
NLopt/OptimizationProblem after run, and if it did not succeed, fail fast or
propagate a clear error instead of computing uMPTP, xzMPTP, and gMPTP from
invalid results. Apply the same safeguard to the matching inverse FORM flow in
SORAAlgorithm so both algorithms handle NLopt failures consistently.
In `@lib/src/RBDOProblem.cxx`:
- Around line 127-132: buildJointDistribution currently ignores the designPoint
argument and always returns a static distribution, so either make the result
depend on the provided [d, p] or remove the parameter everywhere if
design-dependent X is not supported. Update RBDOProblem::buildJointDistribution
and any related contract/doc/callers so the signature and implementation match,
and use designPoint when constructing distributionX_ and the joint
BlockIndependentDistribution.
- Around line 48-119: The parameterized constructor of RBDOProblem copies
distributionZ_ but never marks the problem as having Z, so hasZ_ remains false
and later logic like hasDistributionZ(), buildJointDistribution(), and the
reliability measure path behave as X-only. Fix this by explicitly setting the
Z-present state in the constructor, either by routing the input through
setDistributionZ() or by adding a constructor/flag that updates hasZ_ when
distributionZ is provided. Make sure the change is applied in RBDOProblem and
keeps the existing X/Z joint distribution and constraint setup consistent.
In `@lib/src/RIAAlgorithm.cxx`:
- Around line 136-171: The RIAAlgorithm::operator() currently assumes FORM
always converges and immediately uses FORM::getResult(), which can propagate
invalid reliability values into the outer optimizer. Add failure handling around
the form.run() / form.getResult() sequence by checking the FORMResult
convergence/status (or catching FORM exceptions) and, on failure, return a safe
penalty Point instead of beta - rbdoProblem_.getTargetBeta(). Use the existing
symbols FORM, FORMResult, and RIAAlgorithm::operator() to locate the change.
In `@lib/src/SORAAlgorithm.cxx`:
- Around line 158-232: The shifting vector computed in SORAAlgorithm is
currently dead state because ShiftedConstraintEvaluation only uses xMPTP_ and
zMPTP_ directly and never applies the stored shift. Update the constraint path
so the evaluation uses the shifting vector consistently (for example in
ShiftedConstraintEvaluation::operator() and the constraint setup that feeds
problem.setInequalityConstraint), or remove the shiftingVectors_ computation and
getShiftingVectors() storage if the algorithm is meant to use only MPTP points.
Use the unique symbols ShiftedConstraintEvaluation, shiftingVectors_, and
getShiftingVectors() to keep the behavior consistent.
- Around line 362-367: The shifting vector build in
SORAAlgorithm::compute/update logic is using xzMPTP as if it always matched the
design dimension, which can read past the end and mixes design-space values with
X/Z components. Fix the loop that fills newShift by validating that
activeDesignIndices.getSize() does not exceed the X portion of xzMPTP, and use
only the X components when computing the shift; if the formulation requires lenD
to equal dimX, add an assertion or dimension check near
shiftingVectors_.add(newShift) to enforce that invariant.
In `@python/doc/examples/plot_rbdo_comparison.py`:
- Around line 34-39: The limit-state definition in the plot example uses the
wrong sign in the SymbolicFunction for g, which makes RIA/PMA/SORA enforce the
opposite failure event. Update the constraint expression in the example’s g
definition so it matches the stated form P(Z - d1 - 2d0 ≤ 0), and keep the
ot.Less() convention aligned with that sign. Use the nearby plot_rbdo_comparison
example setup and the SymbolicFunction construction to locate and correct the
expression.
In `@python/doc/examples/plot_rbdo_spur_speed_reducer.py`:
- Around line 28-29: The bounds documentation in the spur speed reducer example
is inconsistent with the implementation in the module’s constraint setup. Update
the source of truth between the LaTeX comment and the bounds used in the example
so they match for d6 and d7; if keeping the literature values, adjust the bounds
in the example code that defines the design variables/constraints, otherwise
revise the LaTeX comment to reflect the current values. Reference the
`plot_rbdo_spur_speed_reducer.py` example and the bounds definition block used
for `d`.
In `@python/src/PMAAlgorithm_doc.i`:
- Around line 10-11: The class docstring for PMAAlgorithm uses the wrong
parameter type for problem; update the documentation in PMAAlgorithm_doc.i so
the problem entry matches the constructor and the setRBDOProblem/getRBDOProblem
APIs, using RBDOProblem instead of RobustOptimizationProblem. Keep the rest of
the docstring consistent with the existing PMAAlgorithm and RBDOProblem
references.
In `@python/src/RBDOProblem_doc.i`:
- Around line 13-26: The RBDOProblem docstring is documenting a non-existent
constructor argument, so update the class documentation to remove
distributionUpdater from the parameter list and delete the misleading note on
distributionX in RBDOProblem_doc.i. Keep the docstring aligned with the actual
RBDOProblem constructor signature and ensure only the real parameters
(distributionX, distributionZ, targetBeta, constraintOperator,
activeDesignIndices, activeParameterIndices, etc.) are described.
In `@python/src/RIAAlgorithm_doc.i`:
- Around line 10-11: The class-level docstring for the RIAAlgorithm parameter
description uses the wrong type name for problem, so update the docstring in
RIAAlgorithm_doc.i to match the actual constructor and accessor signatures.
Replace the RobustOptimizationProblem reference with RBDOProblem in the problem
parameter description, keeping it consistent with the setRBDOProblem and
getRBDOProblem docstrings and the RIAAlgorithm.hxx constructor.
---
Duplicate comments:
In `@lib/src/SORAAlgorithm.cxx`:
- Around line 346-356: The inverse FORM solve in SORAAlgorithm::... uses
NLopt::run() without checking whether the optimization converged. After
invAlgo.run(), inspect the result/status from the inverse problem (as done in
PMAAlgorithm::...) and only use the MPTP and shifting vector when convergence
succeeds; otherwise stop the SORA iteration or propagate an explicit failure so
bad values do not silently drive divergence.
---
Nitpick comments:
In `@lib/src/SORAAlgorithm.cxx`:
- Around line 260-344: Refactor the duplicated inverse FORM setup in
SORAAlgorithm by extracting a shared solver utility that both SORA and PMA can
reuse. The local helper classes PMAObjective and NormConstraint should be moved
into a common implementation alongside the equivalent InverseFormObjective and
SphereConstraint logic, and the repeated NLopt configuration, bounds, starting
point, and iteration-limit code should be centralized into a reusable
solveInverseFORM helper. Ensure SORAAlgorithm::run() calls the shared utility
instead of defining its own near-identical objective/constraint classes.
In `@lib/test/t_RIAAlgorithm_std.cxx`:
- Around line 84-92: The RIA test currently only prints the computed optimum and
does not verify it, so add a tolerance-based assertion in the xStar check path
to validate the expected d1+d2 value against the commented target. Use the
existing ria.getResult().getOptimalPoint() and xStar.getDimension() logic to
locate the spot, and assert that xStar[0] + xStar[1] is close to the expected
2*beta/sqrt(2) value with a reasonable tolerance so the test fails on grossly
incorrect results without being brittle.
In `@python/src/SORAAlgorithm_doc.i`:
- Around line 1-17: The SORAAlgorithm class docstring only describes the
RobustOptimizationProblem constructor, but it also supports an RBDOProblem
constructor. Update the %feature("docstring") OTROBOPT::SORAAlgorithm text to
document the RBDOProblem initialization path alongside the existing one, using
the class name and constructor parameter names so users can see both supported
ways to instantiate SORAAlgorithm.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro Plus
Run ID: d2ce209f-b515-461e-b9b7-ecbee9899984
📒 Files selected for processing (35)
lib/include/otrobopt/OTRobOpt.hxxlib/src/CMakeLists.txtlib/src/PMAAlgorithm.cxxlib/src/RBDOProblem.cxxlib/src/RIAAlgorithm.cxxlib/src/SORAAlgorithm.cxxlib/src/otrobopt/PMAAlgorithm.hxxlib/src/otrobopt/RBDOProblem.hxxlib/src/otrobopt/RIAAlgorithm.hxxlib/src/otrobopt/SORAAlgorithm.hxxlib/test/CMakeLists.txtlib/test/t_MeasureFunction_std.cxxlib/test/t_PMAAlgorithm_std.cxxlib/test/t_RBDOProblem_std.cxxlib/test/t_RIAAlgorithm_std.cxxlib/test/t_RobustOptimizationAlgorithm_std.cxxlib/test/t_RobustOptimizationProblem_std.cxxlib/test/t_SORAAlgorithm_std.cxxpython/doc/examples/plot_rbdo_analytical_case1.pypython/doc/examples/plot_rbdo_comparison.pypython/doc/examples/plot_rbdo_short_column.pypython/doc/examples/plot_rbdo_spur_speed_reducer.pypython/src/CMakeLists.txtpython/src/PMAAlgorithm.ipython/src/PMAAlgorithm_doc.ipython/src/RBDOProblem.ipython/src/RBDOProblem_doc.ipython/src/RIAAlgorithm.ipython/src/RIAAlgorithm_doc.ipython/src/SORAAlgorithm.ipython/src/SORAAlgorithm_doc.ipython/src/otrobopt_module.ipython/test/CMakeLists.txtpython/test/t_RBDOProblemAsOptimizationProblem_std.pypython/test/t_RBDOWithCobyla_std.py
regislebrun
force-pushed
the
RBDO
branch
2 times, most recently
from
July 9, 2026 09:14
a7ed3eb to
01e15bb
Compare
There was a problem hiding this comment.
Actionable comments posted: 6
🧹 Nitpick comments (1)
lib/src/SORAAlgorithm.cxx (1)
344-346: 🚀 Performance & Scalability | 🔵 Trivial | ⚡ Quick winInverse FORM starting point and bounds are not aligned with
beta_target.The starting point
Point(dimXZ, 1.0)has normsqrt(dimXZ), far from the target sphere||u|| = beta_target. FordimXZ=2andbeta_target=3, this is1.41vs3.0. The hardcoded bounds[-7, 7]may also clip the solution for largerbeta_targetvalues.A starting point on the beta-sphere and bounds scaled to
beta_targetwould improve convergence reliability.♻️ Proposed fix
- invFormProb.setBounds(Interval(Point(dimXZ, -7.0), Point(dimXZ, 7.0))); + const Scalar boundMargin = 3.0; + invFormProb.setBounds(Interval(Point(dimXZ, -(targetBeta + boundMargin)), Point(dimXZ, targetBeta + boundMargin))); invAlgo.setProblem(invFormProb); - invAlgo.setStartingPoint(Point(dimXZ, 1.0)); + invAlgo.setStartingPoint(Point(dimXZ, targetBeta / std::sqrt(static_cast<Scalar>(dimXZ)))); invAlgo.setMaximumIterationNumber(100);🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@lib/src/SORAAlgorithm.cxx` around lines 344 - 346, The inverse FORM setup uses a hardcoded starting point and bounds that are not tied to beta_target, so update the invAlgo initialization in SORAAlgorithm to place the starting point on the beta-sphere and derive the invFormProb bounds from beta_target instead of fixed [-7, 7]. Use the existing invFormProb and invAlgo symbols in this block so the initial guess and search region scale with the target reliability level and remain valid for larger beta_target values.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@lib/src/PMAAlgorithm.cxx`:
- Around line 141-146: The parametric reduction in PMA is only fixing active
design variables in gParam, leaving active parameters free and causing a
dimension mismatch when lenP > 0. Update the logic in
PMAAlgorithm::compute/where ParametricFunction is built to combine
rbdoProblem_.getActiveDesignIndices() with
rbdoProblem_.getActiveParameterIndices() before constructing the
ParametricFunction, matching the reduction pattern used in RIAAlgorithm so
inverseTransform’s x+z inputs align with the reduced limit-state.
In `@lib/src/SORAAlgorithm.cxx`:
- Around line 408-425: The SORAAlgorithm persistence logic is missing
startingPoint_, so a save/load cycle loses the user-defined starting point and
run() falls back to the default. Update SORAAlgorithm::save and
SORAAlgorithm::load to serialize and restore startingPoint_ alongside
rbdoProblem_, maxSORAIteration_, soraTolerance_, and mostProbableTargetPoints_,
using the existing Advocate attribute pattern.
- Around line 186-193: The parameter binding in
SORAAlgorithm::computeDesignPoint only fixes the active design variables, so
ParametricFunction still leaves active parameters exposed and later evaluation
can mismatch dimensions. Update the ParametricFunction setup to also bind the
active parameter indices, or explicitly reject non-empty activeParameterIndices
before building gParam. Use the existing symbols activeDesignIndices,
activeParameterIndices, and gParam in SORAAlgorithm to locate and fix the
binding logic.
- Around line 49-57: The SORAAlgorithm constructor that takes a
RobustOptimizationProblem and solver leaves rbdoProblem_ unset, but run() reads
rbdoProblem_ directly, so either initialize rbdoProblem_ in this constructor or
make the precondition explicit by documenting that setRBDOProblem() must be
called before run(); if SORAAlgorithm is intended to be created only from
RBDOProblem, remove or replace this overload to prevent misuse.
In `@python/src/CMakeLists.txt`:
- Line 146: The Sphinx build flags have been weakened by removing the
warning-to-error behavior, which can let documentation issues slip through
unnoticed. Update the SPHINX_FLAGS setting in the CMakeLists logic to preserve
the warning enforcement used by Sphinx, and if a temporary exception is truly
needed for the RBDO docs, fix the underlying warnings in the affected doc
sources or clearly document why the flag change is intentional. Use the
SPHINX_FLAGS configuration block as the anchor for the change.
- Line 38: The warning-suppression flags on `${module_target}` are too broad and
affect all compiled sources, not just SWIG-generated code. Update the CMake
logic in `target_compile_options` for `${module_target}` so the
`-Wno-error=shadow`, `-Wno-error=unused-variable`, and
`-Wno-error=unused-function` settings apply only to SWIG-generated sources,
preferably via `SWIG_COMPILE_FLAGS` or a generator expression tied to those
files. If broad suppression is intentionally kept, add a clear comment near the
`${module_target}` configuration explaining why it is safe.
---
Nitpick comments:
In `@lib/src/SORAAlgorithm.cxx`:
- Around line 344-346: The inverse FORM setup uses a hardcoded starting point
and bounds that are not tied to beta_target, so update the invAlgo
initialization in SORAAlgorithm to place the starting point on the beta-sphere
and derive the invFormProb bounds from beta_target instead of fixed [-7, 7]. Use
the existing invFormProb and invAlgo symbols in this block so the initial guess
and search region scale with the target reliability level and remain valid for
larger beta_target values.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro Plus
Run ID: a0077c2c-8f42-4846-b950-6eb4ad29e1c4
📒 Files selected for processing (49)
.gitignorelib/include/otrobopt/OTRobOpt.hxxlib/src/CMakeLists.txtlib/src/IndividualChanceMeasure.cxxlib/src/JointChanceMeasure.cxxlib/src/MeanMeasure.cxxlib/src/MeanStandardDeviationTradeoffMeasure.cxxlib/src/PMAAlgorithm.cxxlib/src/QuantileMeasure.cxxlib/src/RBDOProblem.cxxlib/src/RIAAlgorithm.cxxlib/src/SORAAlgorithm.cxxlib/src/VarianceMeasure.cxxlib/src/WorstCaseMeasure.cxxlib/src/otrobopt/PMAAlgorithm.hxxlib/src/otrobopt/RBDOProblem.hxxlib/src/otrobopt/RIAAlgorithm.hxxlib/src/otrobopt/SORAAlgorithm.hxxlib/test/CMakeLists.txtlib/test/t_MeasureFunction_std.cxxlib/test/t_PMAAlgorithm_std.cxxlib/test/t_RBDOProblem_std.cxxlib/test/t_RIAAlgorithm_std.cxxlib/test/t_RobustOptimizationAlgorithm_std.cxxlib/test/t_RobustOptimizationProblem_std.cxxlib/test/t_SORAAlgorithm_std.cxxpython/doc/conf.py.inpython/doc/examples/plot_example2.pypython/doc/examples/plot_rbdo_analytical_case1.pypython/doc/examples/plot_rbdo_comparison.pypython/doc/examples/plot_rbdo_short_column.pypython/doc/examples/plot_rbdo_spur_speed_reducer.pypython/doc/user_manual/user_manual.rstpython/src/CMakeLists.txtpython/src/PMAAlgorithm.ipython/src/PMAAlgorithm_doc.ipython/src/RBDOProblem.ipython/src/RBDOProblem_doc.ipython/src/RIAAlgorithm.ipython/src/RIAAlgorithm_doc.ipython/src/SORAAlgorithm.ipython/src/SORAAlgorithm_doc.ipython/src/otrobopt_module.ipython/test/CMakeLists.txtpython/test/sequential_mc.pypython/test/t_RBDOProblemAsOptimizationProblem_std.pypython/test/t_RBDOWithCobyla_std.pypython/test/t_SubsetInverseSampling_R-S.pypython/test/t_SubsetInverseSampling_Waarts_system_series.py
💤 Files with no reviewable changes (2)
- python/doc/examples/plot_example2.py
- python/test/t_SubsetInverseSampling_R-S.py
✅ Files skipped from review due to trivial changes (13)
- lib/test/t_MeasureFunction_std.cxx
- python/src/RIAAlgorithm_doc.i
- python/src/PMAAlgorithm_doc.i
- lib/include/otrobopt/OTRobOpt.hxx
- .gitignore
- python/src/SORAAlgorithm_doc.i
- python/doc/examples/plot_rbdo_comparison.py
- python/doc/user_manual/user_manual.rst
- python/src/RBDOProblem_doc.i
- python/test/t_SubsetInverseSampling_Waarts_system_series.py
- lib/src/WorstCaseMeasure.cxx
- lib/src/JointChanceMeasure.cxx
- lib/src/MeanStandardDeviationTradeoffMeasure.cxx
🚧 Files skipped from review as they are similar to previous changes (22)
- lib/test/t_RIAAlgorithm_std.cxx
- python/src/otrobopt_module.i
- lib/test/t_PMAAlgorithm_std.cxx
- python/test/CMakeLists.txt
- python/src/PMAAlgorithm.i
- lib/test/t_RobustOptimizationAlgorithm_std.cxx
- lib/src/otrobopt/RIAAlgorithm.hxx
- python/src/RBDOProblem.i
- python/src/SORAAlgorithm.i
- python/src/RIAAlgorithm.i
- lib/test/t_RobustOptimizationProblem_std.cxx
- lib/src/otrobopt/PMAAlgorithm.hxx
- lib/test/t_RBDOProblem_std.cxx
- python/doc/examples/plot_rbdo_analytical_case1.py
- python/doc/examples/plot_rbdo_spur_speed_reducer.py
- lib/test/CMakeLists.txt
- lib/src/CMakeLists.txt
- python/test/t_RBDOWithCobyla_std.py
- lib/src/otrobopt/RBDOProblem.hxx
- lib/test/t_SORAAlgorithm_std.cxx
- python/doc/examples/plot_rbdo_short_column.py
- python/test/t_RBDOProblemAsOptimizationProblem_std.py
regislebrun
force-pushed
the
RBDO
branch
2 times, most recently
from
July 9, 2026 10:19
0985d7d to
db3a79f
Compare
There was a problem hiding this comment.
Actionable comments posted: 3
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@lib/src/PMAAlgorithm.cxx`:
- Around line 305-316: PMAAlgorithm::save and PMAAlgorithm::load are missing
persistence for startingPoint_, so a round-trip drops the user-configured
initial point and run() reverts to the solver default. Update both methods to
serialize and restore startingPoint_ alongside rbdoProblem_, following the same
pattern used in SORAAlgorithm. Make sure the member name startingPoint_ is
handled consistently with the existing Advocate saveAttribute/loadAttribute
calls in PMAAlgorithm.
In `@lib/src/RIAAlgorithm.cxx`:
- Line 110: The output description in RIAAlgorithm::setOutputDescription is
inverted relative to the value produced by operator(); update the description
text so it matches the actual returned expression from RIAAlgorithm::operator()
(beta - rbdoProblem_.getTargetBeta()) and keep the naming consistent with the
beta target/output semantics.
- Around line 202-213: The save/load path in RIAAlgorithm is missing persistence
for startingPoint_, so restore it alongside rbdoProblem_ in RIAAlgorithm::save
and RIAAlgorithm::load. Update the serialization in these methods to store and
reload startingPoint_ using the existing Advocate API, keeping the field name
consistent with the class member so the loaded algorithm preserves the
configured initial point instead of falling back to the default.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro Plus
Run ID: d660a04b-8f08-40d0-8014-ec876f3ddf61
📒 Files selected for processing (49)
.gitignorelib/include/otrobopt/OTRobOpt.hxxlib/src/CMakeLists.txtlib/src/IndividualChanceMeasure.cxxlib/src/JointChanceMeasure.cxxlib/src/MeanMeasure.cxxlib/src/MeanStandardDeviationTradeoffMeasure.cxxlib/src/PMAAlgorithm.cxxlib/src/QuantileMeasure.cxxlib/src/RBDOProblem.cxxlib/src/RIAAlgorithm.cxxlib/src/SORAAlgorithm.cxxlib/src/VarianceMeasure.cxxlib/src/WorstCaseMeasure.cxxlib/src/otrobopt/PMAAlgorithm.hxxlib/src/otrobopt/RBDOProblem.hxxlib/src/otrobopt/RIAAlgorithm.hxxlib/src/otrobopt/SORAAlgorithm.hxxlib/test/CMakeLists.txtlib/test/t_MeasureFunction_std.cxxlib/test/t_PMAAlgorithm_std.cxxlib/test/t_RBDOProblem_std.cxxlib/test/t_RIAAlgorithm_std.cxxlib/test/t_RobustOptimizationAlgorithm_std.cxxlib/test/t_RobustOptimizationProblem_std.cxxlib/test/t_SORAAlgorithm_std.cxxpython/doc/conf.py.inpython/doc/examples/plot_example2.pypython/doc/examples/plot_rbdo_analytical_case1.pypython/doc/examples/plot_rbdo_comparison.pypython/doc/examples/plot_rbdo_short_column.pypython/doc/examples/plot_rbdo_spur_speed_reducer.pypython/doc/user_manual/user_manual.rstpython/src/CMakeLists.txtpython/src/PMAAlgorithm.ipython/src/PMAAlgorithm_doc.ipython/src/RBDOProblem.ipython/src/RBDOProblem_doc.ipython/src/RIAAlgorithm.ipython/src/RIAAlgorithm_doc.ipython/src/SORAAlgorithm.ipython/src/SORAAlgorithm_doc.ipython/src/otrobopt_module.ipython/test/CMakeLists.txtpython/test/sequential_mc.pypython/test/t_RBDOProblemAsOptimizationProblem_std.pypython/test/t_RBDOWithCobyla_std.pypython/test/t_SubsetInverseSampling_R-S.pypython/test/t_SubsetInverseSampling_Waarts_system_series.py
💤 Files with no reviewable changes (2)
- python/test/t_SubsetInverseSampling_R-S.py
- python/doc/examples/plot_example2.py
✅ Files skipped from review due to trivial changes (11)
- python/src/RIAAlgorithm_doc.i
- lib/include/otrobopt/OTRobOpt.hxx
- python/doc/user_manual/user_manual.rst
- .gitignore
- python/src/RBDOProblem.i
- lib/src/MeanStandardDeviationTradeoffMeasure.cxx
- python/src/SORAAlgorithm_doc.i
- lib/src/WorstCaseMeasure.cxx
- lib/src/VarianceMeasure.cxx
- python/test/sequential_mc.py
- python/src/PMAAlgorithm_doc.i
🚧 Files skipped from review as they are similar to previous changes (26)
- python/test/CMakeLists.txt
- lib/test/t_MeasureFunction_std.cxx
- python/src/otrobopt_module.i
- python/src/RIAAlgorithm.i
- python/src/PMAAlgorithm.i
- lib/test/t_PMAAlgorithm_std.cxx
- python/doc/conf.py.in
- lib/test/t_RobustOptimizationAlgorithm_std.cxx
- lib/src/CMakeLists.txt
- python/test/t_SubsetInverseSampling_Waarts_system_series.py
- python/src/SORAAlgorithm.i
- lib/test/t_RBDOProblem_std.cxx
- lib/test/t_RobustOptimizationProblem_std.cxx
- lib/test/t_SORAAlgorithm_std.cxx
- lib/src/IndividualChanceMeasure.cxx
- python/test/t_RBDOWithCobyla_std.py
- python/src/RBDOProblem_doc.i
- lib/src/otrobopt/RBDOProblem.hxx
- lib/src/JointChanceMeasure.cxx
- lib/src/otrobopt/SORAAlgorithm.hxx
- lib/src/QuantileMeasure.cxx
- lib/test/CMakeLists.txt
- python/src/CMakeLists.txt
- python/doc/examples/plot_rbdo_short_column.py
- lib/test/t_RIAAlgorithm_std.cxx
- lib/src/MeanMeasure.cxx
jschueller
reviewed
Jul 9, 2026
jschueller
reviewed
Jul 9, 2026
jschueller
reviewed
Jul 9, 2026
jschueller
reviewed
Jul 9, 2026
jschueller
reviewed
Jul 9, 2026
jschueller
reviewed
Jul 9, 2026
drblobfish
reviewed
Jul 10, 2026
drblobfish
left a comment
drblobfish
left a comment
There was a problem hiding this comment.
Hello.
I'm really excited to see all this new work on otrobopt.
I'm not sure if I'm allowed to comment here, but I had some thoughts while reading through this PR and I allowed myself to write them here hoping you might find them useful.
Core classes: - RBDOProblem: define RBDO problems with objective, probabilistic constraint g(d,X,Z), distributions for X and Z, target reliability index, constraint operator (Less/Greater), and design/parameter dimension management - RIAAlgorithm: Reliability Index Approach using FORM to compute reliability indices; outer optimization minimizes f(d,p) subject to beta(d) >= beta_target - PMAAlgorithm: Performance Measure Approach using inverse FORM to compute performance measures for probabilistic constraints - SORAAlgorithm: Sequential Optimization and Reliability Analysis iterating between deterministic optimization with shifted constraints and inverse FORM Thanks to Loic Brevault and Mathieu Balesdent for the RBDO_OT prototype and examples. - sequential_mc.py:140: use self.distributionXi_ instead of global distributionXi (BUG) - plot_example2.py: remove unused import openturns.testing - t_SubsetInverseSampling*.py: remove unused variables (start, CVMC, etc.) - PMAAlgorithm.cxx: remove dead evalLimitState() function - All measure .cxx: rename shadowed Function function() -> func() in operator() methods - RobustOptimizationProblem.hxx: restore needed MeasureEvaluationImplementation.hxx include - SORAAlgorithm.cxx: rename inner xMPTP/zMPTP to xMPTPNew/zMPTPNew to avoid shadow - CMakeLists.txt: enable -Werror=shadow/unused-variable/unused-function for lib; suppress for SWIG wrapper - SPHINX_FLAGS: remove -W to allow pre-existing cross-ref warnings
Member
Author
Hi, Cheers |
There was a problem hiding this comment.
Actionable comments posted: 1
🧹 Nitpick comments (2)
lib/src/SORAAlgorithm.cxx (1)
347-348: 🩺 Stability & Availability | 🔵 Trivial | ⚡ Quick winAborting the whole SORA loop on a single inverse-FORM failure.
A non-convergence in one iteration throws and discards all prior iterations' progress. Consider breaking out and returning the best point found so far (as the RIA path does with a penalty), so a transient inner-solve failure doesn't lose the outer result.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@lib/src/SORAAlgorithm.cxx` around lines 347 - 348, The inverse-FORM failure handling in the SORA loop should not throw and abort the entire algorithm. Update the check on invAlgo.getResult().getStatus() to exit the current SORA iteration path and return the best point accumulated so far, following the existing RIA penalty/fallback behavior while preserving prior iterations’ progress.lib/src/PMAAlgorithm.cxx (1)
150-165: 🚀 Performance & Scalability | 🔵 Trivial | ⚡ Quick winInvariant construction rebuilt on every call.
jointDist,dimXZ,g,activeDesignIndices/activeParameterIndices, andparamIndicesdepend only onrbdoProblem_, which is fixed for the lifetime ofPMAAlgorithm_ConstraintEvaluation. Rebuilding them on everyoperator()call (invoked at least once per outer-optimization iteration, more often with finite-difference gradients) is wasted work; onlyparamValues/gParamgenuinely depend on the currentdesignPoint.♻️ Suggested refactor
PMAAlgorithm_ConstraintEvaluation(const RBDOProblem & rbdoProblem) : EvaluationImplementation() , rbdoProblem_(rbdoProblem) { const UnsignedInteger dim = rbdoProblem.getObjective().getInputDimension(); setInputDescription(Description(dim)); setOutputDescription(Description(1, "-performance_measure")); + jointDist_ = rbdoProblem_.buildJointDistribution(); + g_ = rbdoProblem_.getConstraintFunction(); + activeDesignIndices_ = rbdoProblem_.getActiveDesignIndices(); + activeParameterIndices_ = rbdoProblem_.getActiveParameterIndices(); + paramIndices_ = activeDesignIndices_; + paramIndices_.add(activeParameterIndices_); }🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@lib/src/PMAAlgorithm.cxx` around lines 150 - 165, Cache the invariant objects currently constructed at the start of PMAAlgorithm_ConstraintEvaluation::operator(), including the joint distribution, dimension, constraint function, active index sets, and combined paramIndices, during PMAAlgorithm_ConstraintEvaluation initialization. Update operator() to reuse those cached members and rebuild only paramValues and gParam from the current designPoint.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@lib/src/PMAAlgorithm.cxx`:
- Around line 174-184: Update the sphere minimization setup in the PMA algorithm
to derive the coordinate bounds from targetBeta_ (or the corresponding target
reliability value), ensuring the feasible ||u|| == targetBeta_ sphere is not
excluded for large targets; remove the fixed ±7.0 bounds. Replace the hardcoded
100-iteration cap with the algorithm’s configurable or appropriately scaled
iteration limit, preserving the existing starting-point and optimization flow.
---
Nitpick comments:
In `@lib/src/PMAAlgorithm.cxx`:
- Around line 150-165: Cache the invariant objects currently constructed at the
start of PMAAlgorithm_ConstraintEvaluation::operator(), including the joint
distribution, dimension, constraint function, active index sets, and combined
paramIndices, during PMAAlgorithm_ConstraintEvaluation initialization. Update
operator() to reuse those cached members and rebuild only paramValues and gParam
from the current designPoint.
In `@lib/src/SORAAlgorithm.cxx`:
- Around line 347-348: The inverse-FORM failure handling in the SORA loop should
not throw and abort the entire algorithm. Update the check on
invAlgo.getResult().getStatus() to exit the current SORA iteration path and
return the best point accumulated so far, following the existing RIA
penalty/fallback behavior while preserving prior iterations’ progress.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: defaults
Review profile: CHILL
Plan: Pro Plus
Run ID: 5acccd3f-58fa-4236-83ff-0df5dbead249
📒 Files selected for processing (47)
.gitignorelib/include/otrobopt/OTRobOpt.hxxlib/src/CMakeLists.txtlib/src/IndividualChanceMeasure.cxxlib/src/JointChanceMeasure.cxxlib/src/MeanMeasure.cxxlib/src/MeanStandardDeviationTradeoffMeasure.cxxlib/src/PMAAlgorithm.cxxlib/src/QuantileMeasure.cxxlib/src/RBDOProblem.cxxlib/src/RIAAlgorithm.cxxlib/src/SORAAlgorithm.cxxlib/src/VarianceMeasure.cxxlib/src/WorstCaseMeasure.cxxlib/src/otrobopt/PMAAlgorithm.hxxlib/src/otrobopt/RBDOProblem.hxxlib/src/otrobopt/RIAAlgorithm.hxxlib/src/otrobopt/SORAAlgorithm.hxxlib/test/CMakeLists.txtlib/test/t_MeasureFunction_std.cxxlib/test/t_PMAAlgorithm_std.cxxlib/test/t_RBDOProblem_std.cxxlib/test/t_RIAAlgorithm_std.cxxlib/test/t_RobustOptimizationAlgorithm_std.cxxlib/test/t_RobustOptimizationProblem_std.cxxlib/test/t_SORAAlgorithm_std.cxxpython/doc/examples/plot_example2.pypython/doc/examples/plot_rbdo_analytical_case1.pypython/doc/examples/plot_rbdo_comparison.pypython/doc/examples/plot_rbdo_short_column.pypython/doc/examples/plot_rbdo_spur_speed_reducer.pypython/doc/user_manual/user_manual.rstpython/src/PMAAlgorithm.ipython/src/PMAAlgorithm_doc.ipython/src/RBDOProblem.ipython/src/RBDOProblem_doc.ipython/src/RIAAlgorithm.ipython/src/RIAAlgorithm_doc.ipython/src/SORAAlgorithm.ipython/src/SORAAlgorithm_doc.ipython/src/otrobopt_module.ipython/test/CMakeLists.txtpython/test/sequential_mc.pypython/test/t_RBDOProblemAsOptimizationProblem_std.pypython/test/t_RBDOWithCobyla_std.pypython/test/t_SubsetInverseSampling_R-S.pypython/test/t_SubsetInverseSampling_Waarts_system_series.py
💤 Files with no reviewable changes (3)
- python/doc/examples/plot_example2.py
- .gitignore
- python/test/t_SubsetInverseSampling_R-S.py
🚧 Files skipped from review as they are similar to previous changes (36)
- python/src/RIAAlgorithm_doc.i
- lib/test/t_RBDOProblem_std.cxx
- python/doc/user_manual/user_manual.rst
- python/src/SORAAlgorithm_doc.i
- python/src/RIAAlgorithm.i
- lib/include/otrobopt/OTRobOpt.hxx
- python/test/sequential_mc.py
- python/src/PMAAlgorithm.i
- python/src/RBDOProblem.i
- python/test/CMakeLists.txt
- python/src/SORAAlgorithm.i
- lib/test/t_PMAAlgorithm_std.cxx
- lib/test/t_RobustOptimizationAlgorithm_std.cxx
- lib/src/otrobopt/PMAAlgorithm.hxx
- python/src/RBDOProblem_doc.i
- python/src/PMAAlgorithm_doc.i
- lib/src/VarianceMeasure.cxx
- lib/src/otrobopt/SORAAlgorithm.hxx
- lib/src/QuantileMeasure.cxx
- lib/src/otrobopt/RBDOProblem.hxx
- lib/src/MeanStandardDeviationTradeoffMeasure.cxx
- lib/src/MeanMeasure.cxx
- python/test/t_SubsetInverseSampling_Waarts_system_series.py
- lib/src/IndividualChanceMeasure.cxx
- lib/test/t_RobustOptimizationProblem_std.cxx
- lib/test/t_SORAAlgorithm_std.cxx
- lib/src/WorstCaseMeasure.cxx
- python/test/t_RBDOWithCobyla_std.py
- lib/src/JointChanceMeasure.cxx
- lib/test/CMakeLists.txt
- lib/test/t_RIAAlgorithm_std.cxx
- python/doc/examples/plot_rbdo_analytical_case1.py
- python/doc/examples/plot_rbdo_comparison.py
- python/doc/examples/plot_rbdo_short_column.py
- python/test/t_RBDOProblemAsOptimizationProblem_std.py
- python/doc/examples/plot_rbdo_spur_speed_reducer.py
Comment on lines
+174
to
+184
| OptimizationProblem sphereMinimizationProblem(objFunc); | ||
| sphereMinimizationProblem.setMinimization(true); | ||
| const Function sphereFunc(PMAAlgorithm_SphereConstraint(dimXZ, targetBeta)); | ||
| sphereMinimizationProblem.setEqualityConstraint(sphereFunc); | ||
| sphereMinimizationProblem.setBounds(Interval(Point(dimXZ, -7.0), Point(dimXZ, 7.0))); | ||
| OptimizationAlgorithm algo(OptimizationAlgorithm::Build(sphereMinimizationProblem)); | ||
| if (uStart_.getDimension() != dimXZ) | ||
| uStart_ = Point(dimXZ, 1.0); | ||
| algo.setStartingPoint(uStart_); | ||
| algo.setMaximumIterationNumber(100); | ||
| algo.run(); |
There was a problem hiding this comment.
🎯 Functional Correctness | 🟠 Major | ⚡ Quick win
Hardcoded sphere-search bounds (±7.0) and iteration cap can silently break the constraint for large targetBeta.
setBounds fixes each coordinate of the reduced u-space to [-7, 7] regardless of targetBeta_. The equality constraint requires ||u|| == targetBeta_; if a user sets a target reliability index above ~7 (not unusual for very safety-critical designs), the feasible sphere lies partly or fully outside this box, and the inner sphere minimization (and hence PMA itself) becomes infeasible/silently wrong. The setMaximumIterationNumber(100) on line 183 is the same kind of hardcoded value, though less risky.
🐛 Proposed fix
- sphereMinimizationProblem.setBounds(Interval(Point(dimXZ, -7.0), Point(dimXZ, 7.0)));
+ const Scalar boundMargin = targetBeta + 3.0;
+ sphereMinimizationProblem.setBounds(Interval(Point(dimXZ, -boundMargin), Point(dimXZ, boundMargin)));📝 Committable suggestion
‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.
Suggested change
| OptimizationProblem sphereMinimizationProblem(objFunc); | |
| sphereMinimizationProblem.setMinimization(true); | |
| const Function sphereFunc(PMAAlgorithm_SphereConstraint(dimXZ, targetBeta)); | |
| sphereMinimizationProblem.setEqualityConstraint(sphereFunc); | |
| sphereMinimizationProblem.setBounds(Interval(Point(dimXZ, -7.0), Point(dimXZ, 7.0))); | |
| OptimizationAlgorithm algo(OptimizationAlgorithm::Build(sphereMinimizationProblem)); | |
| if (uStart_.getDimension() != dimXZ) | |
| uStart_ = Point(dimXZ, 1.0); | |
| algo.setStartingPoint(uStart_); | |
| algo.setMaximumIterationNumber(100); | |
| algo.run(); | |
| OptimizationProblem sphereMinimizationProblem(objFunc); | |
| sphereMinimizationProblem.setMinimization(true); | |
| const Function sphereFunc(PMAAlgorithm_SphereConstraint(dimXZ, targetBeta)); | |
| sphereMinimizationProblem.setEqualityConstraint(sphereFunc); | |
| const Scalar boundMargin = targetBeta + 3.0; | |
| sphereMinimizationProblem.setBounds(Interval(Point(dimXZ, -boundMargin), Point(dimXZ, boundMargin))); | |
| OptimizationAlgorithm algo(OptimizationAlgorithm::Build(sphereMinimizationProblem)); | |
| if (uStart_.getDimension() != dimXZ) | |
| uStart_ = Point(dimXZ, 1.0); | |
| algo.setStartingPoint(uStart_); | |
| algo.setMaximumIterationNumber(100); | |
| algo.run(); |
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@lib/src/PMAAlgorithm.cxx` around lines 174 - 184, Update the sphere
minimization setup in the PMA algorithm to derive the coordinate bounds from
targetBeta_ (or the corresponding target reliability value), ensuring the
feasible ||u|| == targetBeta_ sphere is not excluded for large targets; remove
the fixed ±7.0 bounds. Replace the hardcoded 100-iteration cap with the
algorithm’s configurable or appropriately scaled iteration limit, preserving the
existing starting-point and optimization flow.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
https://github.com/l-brevault/RBDO_OT
Core classes:
Thanks to Loic Brevault and Mathieu Balesdent for the RBDO_OT prototype and examples.
Summary by CodeRabbit
New Features
Documentation
Bug Fixes
Tests