1 // @HEADER 2 // ************************************************************************ 3 // 4 // Rapid Optimization Library (ROL) Package 5 // Copyright (2014) Sandia Corporation 6 // 7 // Under terms of Contract DE-AC04-94AL85000, there is a non-exclusive 8 // license for use of this work by or on behalf of the U.S. Government. 9 // 10 // Redistribution and use in source and binary forms, with or without 11 // modification, are permitted provided that the following conditions are 12 // met: 13 // 14 // 1. Redistributions of source code must retain the above copyright 15 // notice, this list of conditions and the following disclaimer. 16 // 17 // 2. Redistributions in binary form must reproduce the above copyright 18 // notice, this list of conditions and the following disclaimer in the 19 // documentation and/or other materials provided with the distribution. 20 // 21 // 3. Neither the name of the Corporation nor the names of the 22 // contributors may be used to endorse or promote products derived from 23 // this software without specific prior written permission. 24 // 25 // THIS SOFTWARE IS PROVIDED BY SANDIA CORPORATION "AS IS" AND ANY 26 // EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 27 // IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 28 // PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL SANDIA CORPORATION OR THE 29 // CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, 30 // EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, 31 // PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR 32 // PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF 33 // LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING 34 // NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS 35 // SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 36 // 37 // Questions? Contact lead developers: 38 // Drew Kouri (dpkouri@sandia.gov) and 39 // Denis Ridzal (dridzal@sandia.gov) 40 // 41 // ************************************************************************ 42 // @HEADER 43 44 #ifndef ROL_BOUND_CONSTRAINT_SIMOPT_H 45 #define ROL_BOUND_CONSTRAINT_SIMOPT_H 46 47 #include "ROL_BoundConstraint.hpp" 48 #include "ROL_Vector_SimOpt.hpp" 49 #include "ROL_Types.hpp" 50 #include <iostream> 51 52 /** @ingroup func_group 53 \class ROL::BoundConstraint 54 \brief Provides the interface to apply upper and lower bound constraints. 55 56 ROL's bound constraint class is to designed to handle point wise bound 57 constraints on optimization variables. That is, let \f$\mathcal{X}\f$ 58 be a Banach space of functions from \f$\Xi\f$ into \f$\mathbb{R}\f$ 59 (for example, \f$\Xi\subset\mathbb{R}^d\f$ for some positive integer \f$d\f$ 60 and \f$\mathcal{X}=L^2(\Xi)\f$ or \f$\Xi = \{1,\ldots,n\}\f$ and 61 \f$\mathcal{X}=\mathbb{R}^n\f$). For any \f$x\in\mathcal{X}\f$, we consider 62 bounds of the form 63 \f[ 64 a(\xi) \le x(\xi) \le b(\xi) \quad \text{for almost every }\xi\in\Xi. 65 \f] 66 Here, \f$a(\xi)\le b(\xi)\f$ for almost every \f$\xi\in\Xi\f$ and \f$a,b\in \mathcal{X}\f$. 67 */ 68 69 70 namespace ROL { 71 72 template <class Real> 73 class BoundConstraint_SimOpt : public BoundConstraint<Real> { 74 private: 75 ROL::Ptr<BoundConstraint<Real> > bnd1_; 76 ROL::Ptr<BoundConstraint<Real> > bnd2_; 77 78 public: ~BoundConstraint_SimOpt()79 ~BoundConstraint_SimOpt() {} 80 81 /** \brief Default constructor. 82 83 The default constructor automatically turns the constraints on. 84 */ BoundConstraint_SimOpt(const ROL::Ptr<BoundConstraint<Real>> & bnd1,const ROL::Ptr<BoundConstraint<Real>> & bnd2)85 BoundConstraint_SimOpt(const ROL::Ptr<BoundConstraint<Real> > &bnd1, 86 const ROL::Ptr<BoundConstraint<Real> > &bnd2) 87 : bnd1_(bnd1), bnd2_(bnd2) { 88 if ( bnd1_->isActivated() || bnd2_->isActivated() ) { 89 BoundConstraint<Real>::activate(); 90 } 91 else { 92 BoundConstraint<Real>::deactivate(); 93 } 94 } 95 96 /** \brief Update bounds. 97 98 The update function allows the user to update the bounds at each new iterations. 99 @param[in] x is the optimization variable. 100 @param[in] flag is set to true if control is changed. 101 @param[in] iter is the outer algorithm iterations count. 102 */ update(const Vector<Real> & x,bool flag=true,int iter=-1)103 void update( const Vector<Real> &x, bool flag = true, int iter = -1 ) { 104 const ROL::Vector_SimOpt<Real> &xs = dynamic_cast<const ROL::Vector_SimOpt<Real>&>( 105 dynamic_cast<const ROL::Vector<Real>&>(x)); 106 if ( bnd1_->isActivated() ) { 107 bnd1_->update(*(xs.get_1()),flag,iter); 108 } 109 if ( bnd2_->isActivated() ) { 110 bnd2_->update(*(xs.get_2()),flag,iter); 111 } 112 } 113 114 /** \brief Project optimization variables onto the bounds. 115 116 This function implements the projection of \f$x\f$ onto the bounds, i.e., 117 \f[ 118 (P_{[a,b]}(x))(\xi) = \min\{b(\xi),\max\{a(\xi),x(\xi)\}\} \quad \text{for almost every }\xi\in\Xi. 119 \f] 120 @param[in,out] x is the optimization variable. 121 */ project(Vector<Real> & x)122 void project( Vector<Real> &x ) { 123 ROL::Vector_SimOpt<Real> &xs = dynamic_cast<ROL::Vector_SimOpt<Real>&>( 124 dynamic_cast<ROL::Vector<Real>&>(x)); 125 if ( bnd1_->isActivated() ) { 126 ROL::Ptr<Vector<Real> > x1 = xs.get_1()->clone(); x1->set(*(xs.get_1())); 127 bnd1_->project(*x1); 128 xs.set_1(*x1); 129 } 130 if ( bnd2_->isActivated() ) { 131 ROL::Ptr<Vector<Real> > x2 = xs.get_2()->clone(); x2->set(*(xs.get_2())); 132 bnd2_->project(*x2); 133 xs.set_2(*x2); 134 } 135 } 136 137 /** \brief Project optimization variables into the interior of the feasible set. 138 139 This function implements the projection of \f$x\f$ into the interior of the 140 feasible set, i.e., 141 \f[ 142 (P_{[a,b]}(x))(\xi) \in (a(\xi),b(\xi)) 143 \quad \text{for almost every }\xi\in\Xi. 144 \f] 145 @param[in,out] x is the optimization variable. 146 */ projectInterior(Vector<Real> & x)147 void projectInterior( Vector<Real> &x ) { 148 ROL::Vector_SimOpt<Real> &xs = dynamic_cast<ROL::Vector_SimOpt<Real>&>( 149 dynamic_cast<ROL::Vector<Real>&>(x)); 150 if ( bnd1_->isActivated() ) { 151 ROL::Ptr<Vector<Real> > x1 = xs.get_1()->clone(); x1->set(*(xs.get_1())); 152 bnd1_->projectInterior(*x1); 153 xs.set_1(*x1); 154 } 155 if ( bnd2_->isActivated() ) { 156 ROL::Ptr<Vector<Real> > x2 = xs.get_2()->clone(); x2->set(*(xs.get_2())); 157 bnd2_->projectInterior(*x2); 158 xs.set_2(*x2); 159 } 160 } 161 162 /** \brief Determine if a vector of Lagrange multipliers is nonnegative components. 163 164 This function returns true if components of \f$l\f$ corresponding to the components of \f$x\f$ 165 that are active at the upper bound are nonpositive or the components of \f$l\f$ corresponding 166 to the components of \f$x\f$ that are active at the lower bound are nonnegative. 167 */ checkMultipliers(const Vector<Real> & l,const Vector<Real> & x)168 bool checkMultipliers( const Vector<Real> &l, const Vector<Real> &x ) { 169 const ROL::Vector_SimOpt<Real> &ls = dynamic_cast<const ROL::Vector_SimOpt<Real>&>( 170 dynamic_cast<const ROL::Vector<Real>&>(l)); 171 const ROL::Vector_SimOpt<Real> &xs = dynamic_cast<const ROL::Vector_SimOpt<Real>&>( 172 dynamic_cast<const ROL::Vector<Real>&>(x)); 173 bool nn1 = true; 174 if ( bnd1_->isActivated() ) { 175 nn1 = bnd1_->checkMultipliers(*(ls.get_1()),*(xs.get_1())); 176 } 177 bool nn2 = true; 178 if ( bnd2_->isActivated() ) { 179 nn2 = bnd2_->checkMultipliers(*(ls.get_2()),*(xs.get_2())); 180 } 181 return (nn1 && nn2); 182 } 183 184 /** \brief Set variables to zero if they correspond to the upper \f$\epsilon\f$-active set. 185 186 This function sets \f$v(\xi)=0\f$ if \f$\xi\in\mathcal{A}^+_\epsilon(x)\f$. Here, 187 the upper \f$\epsilon\f$-active set is defined as 188 \f[ 189 \mathcal{A}^+_\epsilon(x) = \{\,\xi\in\Xi\,:\,x(\xi) = b(\xi)-\epsilon\,\}. 190 \f] 191 @param[out] v is the variable to be pruned. 192 @param[in] x is the current optimization variable. 193 @param[in] eps is the active-set tolerance \f$\epsilon\f$. 194 */ pruneUpperActive(Vector<Real> & v,const Vector<Real> & x,Real eps=0.0)195 void pruneUpperActive( Vector<Real> &v, const Vector<Real> &x, Real eps = 0.0 ) { 196 ROL::Vector_SimOpt<Real> &vs = dynamic_cast<ROL::Vector_SimOpt<Real>&>( 197 dynamic_cast<ROL::Vector<Real>&>(v)); 198 const ROL::Vector_SimOpt<Real> &xs = dynamic_cast<const ROL::Vector_SimOpt<Real>&>( 199 dynamic_cast<const ROL::Vector<Real>&>(x)); 200 if ( bnd1_->isActivated() ) { 201 ROL::Ptr<Vector<Real> > v1 = vs.get_1()->clone(); v1->set(*(vs.get_1())); 202 bnd1_->pruneUpperActive(*v1,*(xs.get_1()),eps); 203 vs.set_1(*v1); 204 } 205 if ( bnd2_->isActivated() ) { 206 ROL::Ptr<Vector<Real> > v2 = vs.get_2()->clone(); v2->set(*(vs.get_2())); 207 bnd2_->pruneUpperActive(*v2,*(xs.get_2()),eps); 208 vs.set_2(*v2); 209 } 210 } 211 212 /** \brief Set variables to zero if they correspond to the upper \f$\epsilon\f$-binding set. 213 214 This function sets \f$v(\xi)=0\f$ if \f$\xi\in\mathcal{B}^+_\epsilon(x)\f$. Here, 215 the upper \f$\epsilon\f$-binding set is defined as 216 \f[ 217 \mathcal{B}^+_\epsilon(x) = \{\,\xi\in\Xi\,:\,x(\xi) = b(\xi)-\epsilon,\; 218 g(\xi) < 0 \,\}. 219 \f] 220 @param[out] v is the variable to be pruned. 221 @param[in] x is the current optimization variable. 222 @param[in] g is the negative search direction. 223 @param[in] eps is the active-set tolerance \f$\epsilon\f$. 224 */ pruneUpperActive(Vector<Real> & v,const Vector<Real> & g,const Vector<Real> & x,Real eps=0.0)225 void pruneUpperActive( Vector<Real> &v, const Vector<Real> &g, const Vector<Real> &x, Real eps = 0.0 ) { 226 ROL::Vector_SimOpt<Real> &vs = dynamic_cast<ROL::Vector_SimOpt<Real>&>( 227 dynamic_cast<ROL::Vector<Real>&>(v)); 228 const ROL::Vector_SimOpt<Real> &gs = dynamic_cast<const ROL::Vector_SimOpt<Real>&>( 229 dynamic_cast<const ROL::Vector<Real>&>(g)); 230 const ROL::Vector_SimOpt<Real> &xs = dynamic_cast<const ROL::Vector_SimOpt<Real>&>( 231 dynamic_cast<const ROL::Vector<Real>&>(x)); 232 if ( bnd1_->isActivated() ) { 233 ROL::Ptr<Vector<Real> > v1 = vs.get_1()->clone(); v1->set(*(vs.get_1())); 234 bnd1_->pruneUpperActive(*v1,*(gs.get_1()),*(xs.get_1()),eps); 235 vs.set_1(*v1); 236 } 237 if ( bnd2_->isActivated() ) { 238 ROL::Ptr<Vector<Real> > v2 = vs.get_2()->clone(); v2->set(*(vs.get_2())); 239 bnd2_->pruneUpperActive(*v2,*(gs.get_2()),*(xs.get_2()),eps); 240 vs.set_2(*v2); 241 } 242 } 243 244 /** \brief Set variables to zero if they correspond to the lower \f$\epsilon\f$-active set. 245 246 This function sets \f$v(\xi)=0\f$ if \f$\xi\in\mathcal{A}^-_\epsilon(x)\f$. Here, 247 the lower \f$\epsilon\f$-active set is defined as 248 \f[ 249 \mathcal{A}^-_\epsilon(x) = \{\,\xi\in\Xi\,:\,x(\xi) = a(\xi)+\epsilon\,\}. 250 \f] 251 @param[out] v is the variable to be pruned. 252 @param[in] x is the current optimization variable. 253 @param[in] eps is the active-set tolerance \f$\epsilon\f$. 254 */ pruneLowerActive(Vector<Real> & v,const Vector<Real> & x,Real eps=0.0)255 void pruneLowerActive( Vector<Real> &v, const Vector<Real> &x, Real eps = 0.0 ) { 256 ROL::Vector_SimOpt<Real> &vs = dynamic_cast<ROL::Vector_SimOpt<Real>&>( 257 dynamic_cast<ROL::Vector<Real>&>(v)); 258 const ROL::Vector_SimOpt<Real> &xs = dynamic_cast<const ROL::Vector_SimOpt<Real>&>( 259 dynamic_cast<const ROL::Vector<Real>&>(x)); 260 if ( bnd1_->isActivated() ) { 261 ROL::Ptr<Vector<Real> > v1 = vs.get_1()->clone(); v1->set(*(vs.get_1())); 262 bnd1_->pruneLowerActive(*v1,*(xs.get_1()),eps); 263 vs.set_1(*v1); 264 } 265 if ( bnd2_->isActivated() ) { 266 ROL::Ptr<Vector<Real> > v2 = vs.get_2()->clone(); v2->set(*(vs.get_2())); 267 bnd2_->pruneLowerActive(*v2,*(xs.get_2()),eps); 268 vs.set_2(*v2); 269 } 270 } 271 272 /** \brief Set variables to zero if they correspond to the lower \f$\epsilon\f$-binding set. 273 274 This function sets \f$v(\xi)=0\f$ if \f$\xi\in\mathcal{B}^-_\epsilon(x)\f$. Here, 275 the lower \f$\epsilon\f$-binding set is defined as 276 \f[ 277 \mathcal{B}^-_\epsilon(x) = \{\,\xi\in\Xi\,:\,x(\xi) = a(\xi)+\epsilon,\; 278 g(\xi) > 0 \,\}. 279 \f] 280 @param[out] v is the variable to be pruned. 281 @param[in] x is the current optimization variable. 282 @param[in] g is the negative search direction. 283 @param[in] eps is the active-set tolerance \f$\epsilon\f$. 284 */ pruneLowerActive(Vector<Real> & v,const Vector<Real> & g,const Vector<Real> & x,Real eps=0.0)285 void pruneLowerActive( Vector<Real> &v, const Vector<Real> &g, const Vector<Real> &x, Real eps = 0.0 ) { 286 ROL::Vector_SimOpt<Real> &vs = dynamic_cast<ROL::Vector_SimOpt<Real>&>( 287 dynamic_cast<ROL::Vector<Real>&>(v)); 288 const ROL::Vector_SimOpt<Real> &gs = dynamic_cast<const ROL::Vector_SimOpt<Real>&>( 289 dynamic_cast<const ROL::Vector<Real>&>(g)); 290 const ROL::Vector_SimOpt<Real> &xs = dynamic_cast<const ROL::Vector_SimOpt<Real>&>( 291 dynamic_cast<const ROL::Vector<Real>&>(x)); 292 if ( bnd1_->isActivated() ) { 293 ROL::Ptr<Vector<Real> > v1 = vs.get_1()->clone(); v1->set(*(vs.get_1())); 294 bnd1_->pruneLowerActive(*v1,*(gs.get_1()),*(xs.get_1()),eps); 295 vs.set_1(*v1); 296 } 297 if ( bnd2_->isActivated() ) { 298 ROL::Ptr<Vector<Real> > v2 = vs.get_2()->clone(); v2->set(*(vs.get_2())); 299 bnd2_->pruneLowerActive(*v2,*(gs.get_2()),*(xs.get_2()),eps); 300 vs.set_2(*v2); 301 } 302 } 303 getLowerBound(void) const304 const ROL::Ptr<const Vector<Real>> getLowerBound( void ) const { 305 const ROL::Ptr<const Vector<Real>> l1 = bnd1_->getLowerBound(); 306 const ROL::Ptr<const Vector<Real>> l2 = bnd2_->getLowerBound(); 307 return ROL::makePtr<Vector_SimOpt<Real>>( ROL::constPtrCast<Vector<Real>>(l1), 308 ROL::constPtrCast<Vector<Real>>(l2) ); 309 } 310 getUpperBound(void) const311 const ROL::Ptr<const Vector<Real>> getUpperBound(void) const { 312 const ROL::Ptr<const Vector<Real>> u1 = bnd1_->getUpperBound(); 313 const ROL::Ptr<const Vector<Real>> u2 = bnd2_->getUpperBound(); 314 return ROL::makePtr<Vector_SimOpt<Real>>( ROL::constPtrCast<Vector<Real>>(u1), 315 ROL::constPtrCast<Vector<Real>>(u2) ); 316 } 317 318 /** \brief Set variables to zero if they correspond to the \f$\epsilon\f$-active set. 319 320 This function sets \f$v(\xi)=0\f$ if \f$\xi\in\mathcal{A}_\epsilon(x)\f$. Here, 321 the \f$\epsilon\f$-active set is defined as 322 \f[ 323 \mathcal{A}_\epsilon(x) = \mathcal{A}^+_\epsilon(x)\cap\mathcal{A}^-_\epsilon(x). 324 \f] 325 @param[out] v is the variable to be pruned. 326 @param[in] x is the current optimization variable. 327 @param[in] eps is the active-set tolerance \f$\epsilon\f$. 328 */ pruneActive(Vector<Real> & v,const Vector<Real> & x,Real eps=0.0)329 void pruneActive( Vector<Real> &v, const Vector<Real> &x, Real eps = 0.0 ) { 330 ROL::Vector_SimOpt<Real> &vs = dynamic_cast<ROL::Vector_SimOpt<Real>&>( 331 dynamic_cast<ROL::Vector<Real>&>(v)); 332 const ROL::Vector_SimOpt<Real> &xs = dynamic_cast<const ROL::Vector_SimOpt<Real>&>( 333 dynamic_cast<const ROL::Vector<Real>&>(x)); 334 if ( bnd1_->isActivated() ) { 335 ROL::Ptr<Vector<Real> > v1 = vs.get_1()->clone(); v1->set(*(vs.get_1())); 336 bnd1_->pruneActive(*v1,*(xs.get_1()),eps); 337 vs.set_1(*v1); 338 } 339 if ( bnd2_->isActivated() ) { 340 ROL::Ptr<Vector<Real> > v2 = vs.get_2()->clone(); v2->set(*(vs.get_2())); 341 bnd2_->pruneActive(*v2,*(xs.get_2()),eps); 342 vs.set_2(*v2); 343 } 344 } 345 346 /** \brief Set variables to zero if they correspond to the \f$\epsilon\f$-binding set. 347 348 This function sets \f$v(\xi)=0\f$ if \f$\xi\in\mathcal{B}_\epsilon(x)\f$. Here, 349 the \f$\epsilon\f$-binding set is defined as 350 \f[ 351 \mathcal{B}^+_\epsilon(x) = \mathcal{B}^+_\epsilon(x)\cap\mathcal{B}^-_\epsilon(x). 352 \f] 353 @param[out] v is the variable to be pruned. 354 @param[in] x is the current optimization variable. 355 @param[in] g is the negative search direction. 356 @param[in] eps is the active-set tolerance \f$\epsilon\f$. 357 */ pruneActive(Vector<Real> & v,const Vector<Real> & g,const Vector<Real> & x,Real eps=0.0)358 void pruneActive( Vector<Real> &v, const Vector<Real> &g, const Vector<Real> &x, Real eps = 0.0 ) { 359 ROL::Vector_SimOpt<Real> &vs = dynamic_cast<ROL::Vector_SimOpt<Real>&>(v); 360 const ROL::Vector_SimOpt<Real> &gs = dynamic_cast<const ROL::Vector_SimOpt<Real>&>(g); 361 const ROL::Vector_SimOpt<Real> &xs = dynamic_cast<const ROL::Vector_SimOpt<Real>&>(x); 362 if ( bnd1_->isActivated() ) { 363 ROL::Ptr<Vector<Real> > v1 = vs.get_1()->clone(); v1->set(*(vs.get_1())); 364 bnd1_->pruneActive(*v1,*(gs.get_1()),*(xs.get_1()),eps); 365 vs.set_1(*v1); 366 } 367 if ( bnd2_->isActivated() ) { 368 ROL::Ptr<Vector<Real> > v2 = vs.get_2()->clone(); v2->set(*(vs.get_2())); 369 bnd2_->pruneActive(*v2,*(gs.get_2()),*(xs.get_2()),eps); 370 vs.set_2(*v2); 371 } 372 } 373 374 /** \brief Check if the vector, v, is feasible. 375 376 This function returns true if \f$v = P_{[a,b]}(v)\f$. 377 @param[in] v is the vector to be checked. 378 */ isFeasible(const Vector<Real> & v)379 bool isFeasible( const Vector<Real> &v ) { 380 const ROL::Vector_SimOpt<Real> &vs = dynamic_cast<const ROL::Vector_SimOpt<Real>&>(v); 381 return (bnd1_->isFeasible(*(vs.get_1()))) && (bnd2_->isFeasible(*(vs.get_2()))); 382 } 383 384 }; // class BoundConstraint 385 386 } // namespace ROL 387 388 #endif 389