|
ROL
|
Provides an interface to run the trust-region algorithm of Lin and More. More...
#include <ROL_TypeB_LinMoreAlgorithm.hpp>
Inheritance diagram for ROL::TypeB::LinMoreAlgorithm< Real >:Public Member Functions | |
| LinMoreAlgorithm (ParameterList &list, const Ptr< Secant< Real >> &secant=nullPtr) | |
| void | run (Vector< Real > &x, const Vector< Real > &g, Objective< Real > &obj, BoundConstraint< Real > &bnd, std::ostream &outStream=std::cout) override |
| Run algorithm on bound constrained problems (Type-B). This general interface supports the use of dual optimization vector spaces, where the user does not define the dual() method. More... | |
| void | writeHeader (std::ostream &os) const override |
| Print iterate header. More... | |
| void | writeName (std::ostream &os) const override |
| Print step name. More... | |
| void | writeOutput (std::ostream &os, bool write_header=false) const override |
| Print iterate status. More... | |
Public Member Functions inherited from ROL::TypeB::Algorithm< Real > | |
| virtual | ~Algorithm () |
| Algorithm () | |
| Constructor, given a step and a status test. More... | |
| void | setStatusTest (const Ptr< StatusTest< Real >> &status, const bool combineStatus=false) |
| virtual void | run (Problem< Real > &problem, std::ostream &outStream=std::cout) |
| Run algorithm on bound constrained problems (Type-B). This is the primary Type-B interface. More... | |
| virtual void | run (Vector< Real > &x, Objective< Real > &obj, BoundConstraint< Real > &bnd, std::ostream &outStream=std::cout) |
| Run algorithm on bound constrained problems (Type-B). This is the primary Type-B interface. More... | |
| virtual void | run (Vector< Real > &x, Objective< Real > &obj, BoundConstraint< Real > &bnd, Constraint< Real > &linear_econ, Vector< Real > &linear_emul, std::ostream &outStream=std::cout) |
| Run algorithm on bound constrained problems with explicit linear constraints (Type-B). This is the primary Type-B with explicit linear constraints interface. More... | |
| virtual void | run (Vector< Real > &x, const Vector< Real > &g, Objective< Real > &obj, BoundConstraint< Real > &bnd, Constraint< Real > &linear_econ, Vector< Real > &linear_emul, const Vector< Real > &linear_eres, std::ostream &outStream=std::cout) |
| Run algorithm on bound constrained problems with explicit linear constraints (Type-B). This general interface supports the use of dual optimization vector spaces, where the user does not define the dual() method. More... | |
| virtual void | run (Vector< Real > &x, Objective< Real > &obj, BoundConstraint< Real > &bnd, Constraint< Real > &linear_icon, Vector< Real > &linear_imul, BoundConstraint< Real > &linear_ibnd, std::ostream &outStream=std::cout) |
| Run algorithm on bound constrained problems with explicit linear constraints (Type-B). This is the primary Type-B with explicit linear constraints interface. More... | |
| virtual void | run (Vector< Real > &x, const Vector< Real > &g, Objective< Real > &obj, BoundConstraint< Real > &bnd, Constraint< Real > &linear_icon, Vector< Real > &linear_imul, BoundConstraint< Real > &linear_ibnd, const Vector< Real > &linear_ires, std::ostream &outStream=std::cout) |
| Run algorithm on bound constrained problems with explicit linear constraints (Type-B). This general interface supports the use of dual optimization vector spaces, where the user does not define the dual() method. More... | |
| virtual void | run (Vector< Real > &x, Objective< Real > &obj, BoundConstraint< Real > &bnd, Constraint< Real > &linear_econ, Vector< Real > &linear_emul, Constraint< Real > &linear_icon, Vector< Real > &linear_imul, BoundConstraint< Real > &linear_ibnd, std::ostream &outStream=std::cout) |
| Run algorithm on bound constrained problems with explicit linear constraints (Type-B). This is the primary Type-B with explicit linear constraints interface. More... | |
| virtual void | run (Vector< Real > &x, const Vector< Real > &g, Objective< Real > &obj, BoundConstraint< Real > &bnd, Constraint< Real > &linear_econ, Vector< Real > &linear_emul, const Vector< Real > &linear_eres, Constraint< Real > &linear_icon, Vector< Real > &linear_imul, BoundConstraint< Real > &linear_ibnd, const Vector< Real > &linear_ires, std::ostream &outStream=std::cout) |
| Run algorithm on bound constrained problems with explicit linear constraints (Type-B). This general interface supports the use of dual optimization vector spaces, where the user does not define the dual() method. More... | |
| virtual void | writeExitStatus (std::ostream &os) const |
| Ptr< const AlgorithmState< Real > > | getState () const |
| void | reset () |
Private Member Functions | |
| void | initialize (Vector< Real > &x, const Vector< Real > &g, Objective< Real > &obj, BoundConstraint< Real > &bnd, std::ostream &outStream=std::cout) |
| Real | dgpstep (Vector< Real > &s, const Vector< Real > &w, const Vector< Real > &x, const Real alpha, std::ostream &outStream=std::cout) const |
| Real | dcauchy (Vector< Real > &s, Real &alpha, Real &q, const Vector< Real > &x, const Vector< Real > &g, const Real del, TrustRegionModel_U< Real > &model, Vector< Real > &dwa, Vector< Real > &dwa1, std::ostream &outStream=std::cout) |
| Real | dprsrch (Vector< Real > &x, Vector< Real > &s, Real &q, const Vector< Real > &g, TrustRegionModel_U< Real > &model, BoundConstraint< Real > &bnd, Vector< Real > &pwa, Vector< Real > &dwa, std::ostream &outStream=std::cout) |
| Real | dtrqsol (const Real xtx, const Real ptp, const Real ptx, const Real del) const |
| Real | dtrpcg (Vector< Real > &w, int &iflag, int &iter, const Vector< Real > &g, const Vector< Real > &x, const Real del, TrustRegionModel_U< Real > &model, BoundConstraint< Real > &bnd, const Real tol, const Real stol, const int itermax, Vector< Real > &p, Vector< Real > &q, Vector< Real > &r, Vector< Real > &t, Vector< Real > &pwa, Vector< Real > &dwa, std::ostream &outStream=std::cout) const |
| void | applyFreeHessian (Vector< Real > &hv, const Vector< Real > &v, const Vector< Real > &x, TrustRegionModel_U< Real > &model, BoundConstraint< Real > &bnd, Real &tol, Vector< Real > &pwa) const |
| void | applyFreePrecond (Vector< Real > &hv, const Vector< Real > &v, const Vector< Real > &x, TrustRegionModel_U< Real > &model, BoundConstraint< Real > &bnd, Real &tol, Vector< Real > &dwa, Vector< Real > &pwa) const |
Private Attributes | |
| Ptr< TrustRegionModel_U< Real > > | model_ |
| Container for trust-region model. More... | |
| Real | delMax_ |
| Maximum trust-region radius (default: ROL_INF) More... | |
| Real | eta0_ |
| Step acceptance threshold (default: 0.05) More... | |
| Real | eta1_ |
| Radius decrease threshold (default: 0.05) More... | |
| Real | eta2_ |
| Radius increase threshold (default: 0.9) More... | |
| Real | gamma0_ |
| Radius decrease rate (negative rho) (default: 0.0625) More... | |
| Real | gamma1_ |
| Radius decrease rate (positive rho) (default: 0.25) More... | |
| Real | gamma2_ |
| Radius increase rate (default: 2.5) More... | |
| Real | TRsafe_ |
| Safeguard size for numerically evaluating ratio (default: 1e2) More... | |
| Real | eps_ |
| Safeguard for numerically evaluating ratio. More... | |
| bool | interpRad_ |
| Interpolate the trust-region radius if ratio is negative (default: false) More... | |
| TRUtils::ETRFlag | TRflag_ |
| Trust-region exit flag. More... | |
| int | SPflag_ |
| Subproblem solver termination flag. More... | |
| int | SPiter_ |
| Subproblem solver iteration count. More... | |
| ESecant | esec_ |
| Secant type (default: Limited-Memory BFGS) More... | |
| bool | useSecantPrecond_ |
| Flag to use secant as a preconditioner (default: false) More... | |
| bool | useSecantHessVec_ |
| Flag to use secant as Hessian (default: false) More... | |
| Real | tol1_ |
| Absolute tolerance for truncated CG (default: 1e-4) More... | |
| Real | tol2_ |
| Relative tolerance for truncated CG (default: 1e-2) More... | |
| int | maxit_ |
| Maximum number of CG iterations (default: 20) More... | |
| int | minit_ |
| Maximum number of minor (subproblem solve) iterations (default: 10) More... | |
| Real | mu0_ |
| Sufficient decrease parameter (default: 1e-2) More... | |
| Real | spexp_ |
| Relative tolerance exponent for subproblem solve (default: 1, range: [1,2]) More... | |
| int | redlim_ |
| Maximum number of Cauchy point reduction steps (default: 10) More... | |
| int | explim_ |
| Maximum number of Cauchy point expansion steps (default: 10) More... | |
| Real | alpha_ |
| Initial Cauchy point step length (default: 1.0) More... | |
| bool | normAlpha_ |
| Normalize initial Cauchy point step length (default: false) More... | |
| Real | interpf_ |
| Backtracking rate for Cauchy point computation (default: 1e-1) More... | |
| Real | extrapf_ |
| Extrapolation rate for Cauchy point computation (default: 1e1) More... | |
| Real | qtol_ |
| Relative tolerance for computed decrease in Cauchy point computation (default: 1-8) More... | |
| Real | interpfPS_ |
| Backtracking rate for projected search (default: 0.5) More... | |
| int | pslim_ |
| Maximum number of projected search steps (default: 20) More... | |
| int | nhess_ |
| Number of Hessian applications. More... | |
| unsigned | verbosity_ |
| Output level (default: 0) More... | |
| bool | writeHeader_ |
| Flag to write header at every iteration. More... | |
| bool | hasEcon_ |
| Flag signifies if equality constraints exist. More... | |
| Ptr< ReducedLinearConstraint< Real > > | rcon_ |
| Equality constraint restricted to current active variables. More... | |
| Ptr< NullSpaceOperator< Real > > | ns_ |
| Null space projection onto reduced equality constraint Jacobian. More... | |
Additional Inherited Members | |
Protected Member Functions inherited from ROL::TypeB::Algorithm< Real > | |
| void | initialize (const Vector< Real > &x, const Vector< Real > &g) |
| Real | optimalityCriterion (const Vector< Real > &x, const Vector< Real > &g, Vector< Real > &primal, std::ostream &outStream=std::cout) const |
Protected Attributes inherited from ROL::TypeB::Algorithm< Real > | |
| const Ptr< CombinedStatusTest< Real > > | status_ |
| const Ptr< AlgorithmState< Real > > | state_ |
| Ptr< PolyhedralProjection< Real > > | proj_ |
Provides an interface to run the trust-region algorithm of Lin and More.
Definition at line 61 of file ROL_TypeB_LinMoreAlgorithm.hpp.
| ROL::TypeB::LinMoreAlgorithm< Real >::LinMoreAlgorithm | ( | ParameterList & | list, |
| const Ptr< Secant< Real >> & | secant = nullPtr |
||
| ) |
Definition at line 51 of file ROL_TypeB_LinMoreAlgorithm_Def.hpp.
References ROL::SECANTMODE_BOTH, ROL::SECANTMODE_FORWARD, ROL::SECANTMODE_INVERSE, and ROL::StringToESecant().
|
overridevirtual |
Run algorithm on bound constrained problems (Type-B). This general interface supports the use of dual optimization vector spaces, where the user does not define the dual() method.
Implements ROL::TypeB::Algorithm< Real >.
Definition at line 154 of file ROL_TypeB_LinMoreAlgorithm_Def.hpp.
References ROL::Accept, ROL::Vector< Real >::clone(), ROL::Vector< Real >::dual(), ROL::Objective< Real >::gradient(), ROL::TRUtils::NPOSPREDNEG, ROL::TypeB::Algorithm< Real >::optimalityCriterion(), ROL::Vector< Real >::plus(), ROL::TRUtils::POSPREDNEG, ROL::BoundConstraint< Real >::pruneActive(), ROL::Revert, ROL::Vector< Real >::set(), ROL::TRUtils::SUCCESS, ROL::Trial, ROL::TRUtils::TRNAN, ROL::Objective< Real >::update(), ROL::Objective< Real >::value(), ROL::TypeB::Algorithm< Real >::writeExitStatus(), and zero.
Referenced by main().
|
overridevirtual |
Print iterate header.
Reimplemented from ROL::TypeB::Algorithm< Real >.
Definition at line 620 of file ROL_TypeB_LinMoreAlgorithm_Def.hpp.
References ROL::CG_FLAG_SUCCESS, ROL::CG_FLAG_UNDEFINED, ROL::ECGFlagToString(), ROL::TRUtils::ETRFlagToString(), ROL::NumberToString(), ROL::TRUtils::SUCCESS, and ROL::TRUtils::UNDEFINED.
|
overridevirtual |
Print step name.
Reimplemented from ROL::TypeB::Algorithm< Real >.
Definition at line 671 of file ROL_TypeB_LinMoreAlgorithm_Def.hpp.
|
overridevirtual |
Print iterate status.
Reimplemented from ROL::TypeB::Algorithm< Real >.
Definition at line 678 of file ROL_TypeB_LinMoreAlgorithm_Def.hpp.
|
private |
Definition at line 106 of file ROL_TypeB_LinMoreAlgorithm_Def.hpp.
References ROL::Objective< Real >::gradient(), ROL::Initial, ROL::TypeB::Algorithm< Real >::initialize(), ROL::Objective< Real >::update(), and ROL::Objective< Real >::value().
|
private |
Definition at line 328 of file ROL_TypeB_LinMoreAlgorithm_Def.hpp.
References ROL::Vector< Real >::axpy(), ROL::Vector< Real >::norm(), and ROL::Vector< Real >::set().
|
private |
Definition at line 338 of file ROL_TypeB_LinMoreAlgorithm_Def.hpp.
References ROL::Vector< Real >::apply(), ROL::Vector< Real >::dot(), ROL::TrustRegionModel_U< Real >::hessVec(), and ROL::Vector< Real >::set().
|
private |
Definition at line 427 of file ROL_TypeB_LinMoreAlgorithm_Def.hpp.
References ROL::Vector< Real >::apply(), ROL::Vector< Real >::dot(), ROL::TrustRegionModel_U< Real >::hessVec(), ROL::Vector< Real >::plus(), and ROL::Vector< Real >::set().
|
private |
Definition at line 467 of file ROL_TypeB_LinMoreAlgorithm_Def.hpp.
References zero.
|
private |
Definition at line 489 of file ROL_TypeB_LinMoreAlgorithm_Def.hpp.
References ROL::Vector< Real >::apply(), ROL::Vector< Real >::axpy(), ROL::Vector< Real >::dot(), ROL::Vector< Real >::norm(), ROL::Vector< Real >::plus(), ROL::Vector< Real >::scale(), ROL::Vector< Real >::set(), zero, and ROL::Vector< Real >::zero().
|
private |
Definition at line 569 of file ROL_TypeB_LinMoreAlgorithm_Def.hpp.
References ROL::Vector< Real >::dual(), ROL::TrustRegionModel_U< Real >::hessVec(), ROL::BoundConstraint< Real >::pruneActive(), ROL::Vector< Real >::set(), and zero.
|
private |
Definition at line 590 of file ROL_TypeB_LinMoreAlgorithm_Def.hpp.
References ROL::Vector< Real >::dual(), ROL::TrustRegionModel_U< Real >::precond(), ROL::BoundConstraint< Real >::pruneActive(), ROL::Vector< Real >::set(), and zero.
|
private |
Container for trust-region model.
Definition at line 63 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Maximum trust-region radius (default: ROL_INF)
Definition at line 66 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Step acceptance threshold (default: 0.05)
Definition at line 67 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Radius decrease threshold (default: 0.05)
Definition at line 68 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Radius increase threshold (default: 0.9)
Definition at line 69 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Radius decrease rate (negative rho) (default: 0.0625)
Definition at line 70 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Radius decrease rate (positive rho) (default: 0.25)
Definition at line 71 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Radius increase rate (default: 2.5)
Definition at line 72 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Safeguard size for numerically evaluating ratio (default: 1e2)
Definition at line 73 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Safeguard for numerically evaluating ratio.
Definition at line 74 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Interpolate the trust-region radius if ratio is negative (default: false)
Definition at line 75 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Trust-region exit flag.
Definition at line 78 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Subproblem solver termination flag.
Definition at line 79 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Subproblem solver iteration count.
Definition at line 80 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Secant type (default: Limited-Memory BFGS)
Definition at line 83 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Flag to use secant as a preconditioner (default: false)
Definition at line 84 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Flag to use secant as Hessian (default: false)
Definition at line 85 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Absolute tolerance for truncated CG (default: 1e-4)
Definition at line 88 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Relative tolerance for truncated CG (default: 1e-2)
Definition at line 89 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Maximum number of CG iterations (default: 20)
Definition at line 90 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Maximum number of minor (subproblem solve) iterations (default: 10)
Definition at line 93 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Sufficient decrease parameter (default: 1e-2)
Definition at line 94 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Relative tolerance exponent for subproblem solve (default: 1, range: [1,2])
Definition at line 95 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Maximum number of Cauchy point reduction steps (default: 10)
Definition at line 96 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Maximum number of Cauchy point expansion steps (default: 10)
Definition at line 97 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Initial Cauchy point step length (default: 1.0)
Definition at line 98 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Normalize initial Cauchy point step length (default: false)
Definition at line 99 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Backtracking rate for Cauchy point computation (default: 1e-1)
Definition at line 100 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Extrapolation rate for Cauchy point computation (default: 1e1)
Definition at line 101 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Relative tolerance for computed decrease in Cauchy point computation (default: 1-8)
Definition at line 102 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Backtracking rate for projected search (default: 0.5)
Definition at line 103 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Maximum number of projected search steps (default: 20)
Definition at line 104 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
mutableprivate |
Number of Hessian applications.
Definition at line 106 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Output level (default: 0)
Definition at line 107 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Flag to write header at every iteration.
Definition at line 108 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Flag signifies if equality constraints exist.
Definition at line 110 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Equality constraint restricted to current active variables.
Definition at line 111 of file ROL_TypeB_LinMoreAlgorithm.hpp.
|
private |
Null space projection onto reduced equality constraint Jacobian.
Definition at line 112 of file ROL_TypeB_LinMoreAlgorithm.hpp.