ROL
ROL_QuadraticPenalty.hpp
Go to the documentation of this file.
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_QUADRATICPENALTY_H
45 #define ROL_QUADRATICPENALTY_H
46 
47 #include "ROL_Objective.hpp"
48 #include "ROL_Constraint.hpp"
49 #include "ROL_Vector.hpp"
50 #include "ROL_Types.hpp"
51 #include "ROL_Ptr.hpp"
52 #include <iostream>
53 
81 namespace ROL {
82 
83 template <class Real>
84 class QuadraticPenalty : public Objective<Real> {
85 private:
86  // Required for quadratic penalty definition
87  const ROL::Ptr<Constraint<Real> > con_;
88  ROL::Ptr<Vector<Real> > multiplier_;
90 
91  // Auxiliary storage
92  ROL::Ptr<Vector<Real> > primalMultiplierVector_;
93  ROL::Ptr<Vector<Real> > dualOptVector_;
94  ROL::Ptr<Vector<Real> > primalConVector_;
95 
96  // Constraint evaluations
97  ROL::Ptr<Vector<Real> > conValue_;
98  Real cscale_;
99 
100  // Evaluation counters
101  int ncval_;
102 
103  // User defined options
104  const bool useScaling_;
105  const int HessianApprox_;
106 
107  // Flags to recompute quantities
109 
110  void evaluateConstraint(const Vector<Real> &x, Real &tol) {
111  if ( !isConstraintComputed_ ) {
112  // Evaluate constraint
113  con_->value(*conValue_,x,tol); ncval_++;
114  isConstraintComputed_ = true;
115  }
116  }
117 
118 public:
119  QuadraticPenalty(const ROL::Ptr<Constraint<Real> > &con,
120  const Vector<Real> &multiplier,
121  const Real penaltyParameter,
122  const Vector<Real> &optVec,
123  const Vector<Real> &conVec,
124  const bool useScaling = false,
125  const int HessianApprox = 0 )
126  : con_(con), penaltyParameter_(penaltyParameter), cscale_(1), ncval_(0),
127  useScaling_(useScaling), HessianApprox_(HessianApprox), isConstraintComputed_(false) {
128 
129  dualOptVector_ = optVec.dual().clone();
130  primalConVector_ = conVec.clone();
131  conValue_ = conVec.clone();
132  multiplier_ = multiplier.clone();
133  primalMultiplierVector_ = multiplier.clone();
134  }
135 
136  void setScaling(const Real cscale = 1) {
137  cscale_ = cscale;
138  }
139 
140  virtual void update( const Vector<Real> &x, bool flag = true, int iter = -1 ) {
141  con_->update(x,flag,iter);
142  isConstraintComputed_ = ( flag ? false : isConstraintComputed_ );
143  }
144 
145  virtual Real value( const Vector<Real> &x, Real &tol ) {
146  // Evaluate constraint
147  evaluateConstraint(x,tol);
148  // Apply Lagrange multiplier to constraint
149  Real cval = cscale_*multiplier_->dot(conValue_->dual());
150  // Compute penalty term
151  Real pval = cscale_*cscale_*conValue_->dot(*conValue_);
152  // Compute quadratic penalty value
153  const Real half(0.5);
154  Real val(0);
155  if (useScaling_) {
156  val = cval/penaltyParameter_ + half*pval;
157  }
158  else {
159  val = cval + half*penaltyParameter_*pval;
160  }
161  return val;
162  }
163 
164  virtual void gradient( Vector<Real> &g, const Vector<Real> &x, Real &tol ) {
165  // Evaluate constraint
166  evaluateConstraint(x,tol);
167  // Compute gradient of Augmented Lagrangian
168  primalMultiplierVector_->set(conValue_->dual());
169  if ( useScaling_ ) {
172  }
173  else {
176  }
177  con_->applyAdjointJacobian(g,*primalMultiplierVector_,x,tol);
178  }
179 
180  virtual void hessVec( Vector<Real> &hv, const Vector<Real> &v, const Vector<Real> &x, Real &tol ) {
181  // Apply objective Hessian to a vector
182  if (HessianApprox_ < 3) {
183  con_->applyJacobian(*primalConVector_,v,x,tol);
184  con_->applyAdjointJacobian(hv,primalConVector_->dual(),x,tol);
185  if (!useScaling_) {
187  }
188  else {
189  hv.scale(cscale_*cscale_);
190  }
191 
192  if (HessianApprox_ == 1) {
193  // Apply Augmented Lagrangian Hessian to a vector
195  if ( useScaling_ ) {
197  }
198  else {
200  }
201  con_->applyAdjointHessian(*dualOptVector_,*primalMultiplierVector_,v,x,tol);
202  hv.plus(*dualOptVector_);
203  }
204 
205  if (HessianApprox_ == 0) {
206  // Evaluate constraint
207  evaluateConstraint(x,tol);
208  // Apply Augmented Lagrangian Hessian to a vector
209  primalMultiplierVector_->set(conValue_->dual());
210  if ( useScaling_ ) {
213  }
214  else {
217  }
218  con_->applyAdjointHessian(*dualOptVector_,*primalMultiplierVector_,v,x,tol);
219  hv.plus(*dualOptVector_);
220  }
221  }
222  else {
223  hv.zero();
224  }
225  }
226 
227  // Return constraint value
228  virtual void getConstraintVec(Vector<Real> &c, const Vector<Real> &x) {
229  Real tol = std::sqrt(ROL_EPSILON<Real>());
230  // Evaluate constraint
231  evaluateConstraint(x,tol);
232  c.set(*conValue_);
233  }
234 
235  // Return total number of constraint evaluations
236  virtual int getNumberConstraintEvaluations(void) const {
237  return ncval_;
238  }
239 
240  // Reset with upated penalty parameter
241  virtual void reset(const Vector<Real> &multiplier, const Real penaltyParameter) {
242  ncval_ = 0;
243  multiplier_->set(multiplier);
244  penaltyParameter_ = penaltyParameter;
245  }
246 }; // class AugmentedLagrangian
247 
248 } // namespace ROL
249 
250 #endif
Provides the interface to evaluate objective functions.
virtual const Vector & dual() const
Return dual representation of , for example, the result of applying a Riesz map, or change of basis...
Definition: ROL_Vector.hpp:226
virtual void scale(const Real alpha)=0
Compute where .
virtual void reset(const Vector< Real > &multiplier, const Real penaltyParameter)
virtual ROL::Ptr< Vector > clone() const =0
Clone to make a new (uninitialized) vector.
virtual void hessVec(Vector< Real > &hv, const Vector< Real > &v, const Vector< Real > &x, Real &tol)
Apply Hessian approximation to vector.
virtual void plus(const Vector &x)=0
Compute , where .
ROL::Ptr< Vector< Real > > primalConVector_
QuadraticPenalty(const ROL::Ptr< Constraint< Real > > &con, const Vector< Real > &multiplier, const Real penaltyParameter, const Vector< Real > &optVec, const Vector< Real > &conVec, const bool useScaling=false, const int HessianApprox=0)
Provides the interface to evaluate the quadratic constraint penalty.
Contains definitions of custom data types in ROL.
virtual void zero()
Set to zero vector.
Definition: ROL_Vector.hpp:167
Defines the linear algebra or vector space interface.
Definition: ROL_Vector.hpp:80
const ROL::Ptr< Constraint< Real > > con_
void evaluateConstraint(const Vector< Real > &x, Real &tol)
virtual void gradient(Vector< Real > &g, const Vector< Real > &x, Real &tol)
Compute gradient.
ROL::Ptr< Vector< Real > > multiplier_
virtual int getNumberConstraintEvaluations(void) const
virtual Real value(const Vector< Real > &x, Real &tol)
Compute value.
void setScaling(const Real cscale=1)
ROL::Ptr< Vector< Real > > conValue_
virtual void set(const Vector &x)
Set where .
Definition: ROL_Vector.hpp:209
virtual void update(const Vector< Real > &x, bool flag=true, int iter=-1)
Update objective function.
virtual void getConstraintVec(Vector< Real > &c, const Vector< Real > &x)
ROL::Ptr< Vector< Real > > primalMultiplierVector_
Defines the general constraint operator interface.
ROL::Ptr< Vector< Real > > dualOptVector_