ROL
ROL_PD_MeanSemiDeviation.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_PD_MEANSEMIDEVIATION_HPP
45#define ROL_PD_MEANSEMIDEVIATION_HPP
46
48
49namespace ROL {
50
51template<class Real>
53private:
54 Real coeff_;
55
56 Ptr<ScalarController<Real>> values_;
57 Ptr<ScalarController<Real>> gradvecs_;
58 Ptr<VectorController<Real>> gradients_;
59 Ptr<VectorController<Real>> hessvecs_;
60
61 using RandVarFunctional<Real>::val_;
62 using RandVarFunctional<Real>::g_;
63 using RandVarFunctional<Real>::gv_;
64 using RandVarFunctional<Real>::hv_;
66
67 using RandVarFunctional<Real>::point_;
69
74
78 using PD_RandVarFunctional<Real>::ppf;
79
80 void initializeStorage(void) {
81 values_ = makePtr<ScalarController<Real>>();
82 gradvecs_ = makePtr<ScalarController<Real>>();
83 gradients_ = makePtr<VectorController<Real>>();
84 hessvecs_ = makePtr<VectorController<Real>>();
85
88 }
89
90 void clear(void) {
91 gradvecs_->reset();
92 hessvecs_->reset();
93 }
94
95 void checkInputs(void) {
96 Real zero(0);
97 ROL_TEST_FOR_EXCEPTION((coeff_ < zero), std::invalid_argument,
98 ">>> ERROR (ROL::PD_MeanSemiDeviation): Element of coefficient array out of range!");
100 }
101
102public:
103 PD_MeanSemiDeviation(const Real coeff)
104 : PD_RandVarFunctional<Real>(), coeff_(coeff) {
105 checkInputs();
106 }
107
108 void setStorage(const Ptr<ScalarController<Real>> &value_storage,
109 const Ptr<VectorController<Real>> &gradient_storage) {
110 values_ = value_storage;
111 gradients_ = gradient_storage;
113 }
114
115 void setHessVecStorage(const Ptr<ScalarController<Real>> &gradvec_storage,
116 const Ptr<VectorController<Real>> &hessvec_storage) {
117 gradvecs_ = gradvec_storage;
118 hessvecs_ = hessvec_storage;
120 }
121
122 void initialize(const Vector<Real> &x) {
124 clear();
125 }
126
128 const Vector<Real> &x,
129 const std::vector<Real> &xstat,
130 Real &tol) {
131 Real val = computeValue(obj,x,tol);
132 val_ += weight_ * val;
133 }
134
135 Real getValue(const Vector<Real> &x,
136 const std::vector<Real> &xstat,
137 SampleGenerator<Real> &sampler) {
138 // Compute expected value
139 Real ev(0);
140 sampler.sumAll(&val_,&ev,1);
141 // Compute deviation
142 Real diff(0), pf0(0), dev(0), weight(0), lam(0);
143 for (int i = sampler.start(); i < sampler.numMySamples(); ++i) {
144 values_->get(diff,sampler.getMyPoint(i));
145 diff -= ev;
146 setValue(diff, sampler.getMyPoint(i));
147 getMultiplier(lam,sampler.getMyPoint(i));
148 weight = sampler.getMyWeight(i);
149 pf0 += weight * ppf(diff, lam, getPenaltyParameter(), 0);
150 }
151 sampler.sumAll(&pf0,&dev,1);
152 dev *= coeff_;
153 // Return mean plus deviation
154 return ev + dev;
155 }
156
158 const Vector<Real> &x,
159 const std::vector<Real> &xstat,
160 Real &tol) {
161 Real val = computeValue(obj,x,tol);
162 val_ += weight_ * val;
163 computeGradient(*dualVector_,obj,x,tol);
164 g_->axpy(weight_,*dualVector_);
165 }
166
168 std::vector<Real> &gstat,
169 const Vector<Real> &x,
170 const std::vector<Real> &xstat,
171 SampleGenerator<Real> &sampler) {
172 // Compute expected value
173 Real ev(0);
174 sampler.sumAll(&val_,&ev,1);
175 // Compute deviation
176 hv_->zero(); dualVector_->zero();
177 Real diff(0), pf(0), pf1(0), dev(0), one(1), weight(0), lam(0);
178 for (int i = sampler.start(); i < sampler.numMySamples(); ++i) {
179 values_->get(diff,sampler.getMyPoint(i));
180 diff -= ev;
181 getMultiplier(lam,sampler.getMyPoint(i));
182 weight = sampler.getMyWeight(i);
183 pf1 = weight * ppf(diff, lam, getPenaltyParameter(), 1);
184 pf += pf1;
185 gradients_->get(*hv_, sampler.getMyPoint(i));
186 dualVector_->axpy(coeff_ * pf1, *hv_);
187 }
188 sampler.sumAll(&pf,&dev,1);
189 g_->scale(one - coeff_ * dev);
190 g_->plus(*dualVector_);
191 sampler.sumAll(*g_,g);
192 }
193
195 const Vector<Real> &v,
196 const std::vector<Real> &vstat,
197 const Vector<Real> &x,
198 const std::vector<Real> &xstat,
199 Real &tol) {
200 Real val = computeValue(obj,x,tol);
201 val_ += weight_ * val;
202 Real gv = computeGradVec(*dualVector_,obj,v,x,tol);
203 gv_ += weight_ * gv;
204 g_->axpy(weight_, *dualVector_);
205 computeHessVec(*dualVector_,obj,v,x,tol);
206 hv_->axpy(weight_, *dualVector_);
207 }
208
210 std::vector<Real> &hvstat,
211 const Vector<Real> &v,
212 const std::vector<Real> &vstat,
213 const Vector<Real> &x,
214 const std::vector<Real> &xstat,
215 SampleGenerator<Real> &sampler) {
216 const Real one(1);
217 // Compute expected value
218 std::vector<Real> mval(2), gval(2);
219 mval[0] = val_; mval[1] = gv_;
220 sampler.sumAll(&mval[0],&gval[0],2);
221 Real ev = gval[0], egv = gval[1];
222 // Compute deviation
223 std::vector<Real> mvec(3), gvec(3);
224 Real diff(0), gv(0), pf1(0), pf2(0), weight(0), lam(0);
225 for (int i = sampler.start(); i < sampler.numMySamples(); ++i) {
226 values_->get(diff,sampler.getMyPoint(i));
227 gradvecs_->get(gv,sampler.getMyPoint(i));
228 getMultiplier(lam,sampler.getMyPoint(i));
229 weight = sampler.getMyWeight(i);
230 diff -= ev;
231 pf1 = ppf(diff, lam, getPenaltyParameter(), 1);
232 pf2 = ppf(diff, lam, getPenaltyParameter(), 2);
233 mvec[0] += weight * pf1;
234 mvec[1] += weight * pf2;
235 mvec[2] += weight * pf2 * gv;
236 }
237 sampler.sumAll(&mvec[0],&gvec[0],3);
238 Real c1 = one - coeff_ * gvec[0];
239 Real c2 = coeff_ * (gvec[1]*egv - gvec[2]);
240 hv_->scale(c1);
241 hv_->axpy(c2, *g_);
242 sampler.sumAll(*hv_,hv);
243
244 dualVector_->zero(); hv_->zero(); g_->zero();
245 for (int i = sampler.start(); i < sampler.numMySamples(); ++i) {
246 values_->get(diff,sampler.getMyPoint(i));
247 gradients_->get(*g_,sampler.getMyPoint(i));
248 gradvecs_->get(gv,sampler.getMyPoint(i));
249 hessvecs_->get(*dualVector_,sampler.getMyPoint(i));
250 getMultiplier(lam,sampler.getMyPoint(i));
251 weight = sampler.getMyWeight(i);
252 diff -= ev;
253 pf1 = ppf(diff, lam, getPenaltyParameter(), 1);
254 pf2 = ppf(diff, lam, getPenaltyParameter(), 2);
255 hv_->axpy(weight * coeff_ * pf2 * (gv-egv), *g_);
256 hv_->axpy(weight * coeff_ *pf1, *dualVector_);
257 }
258 sampler.sumAll(*hv_, *dualVector_);
259 hv.plus(*dualVector_);
260 }
261};
262
263}
264
265#endif
Objective_SerialSimOpt(const Ptr< Obj > &obj, const V &ui) z0_ zero()
Provides the interface to evaluate objective functions.
void setStorage(const Ptr< ScalarController< Real > > &value_storage, const Ptr< VectorController< Real > > &gradient_storage)
void updateHessVec(Objective< Real > &obj, const Vector< Real > &v, const std::vector< Real > &vstat, const Vector< Real > &x, const std::vector< Real > &xstat, Real &tol)
Update internal risk measure storage for Hessian-time-a-vector computation.
void updateGradient(Objective< Real > &obj, const Vector< Real > &x, const std::vector< Real > &xstat, Real &tol)
Update internal risk measure storage for gradient computation.
Ptr< VectorController< Real > > gradients_
void initialize(const Vector< Real > &x)
Initialize temporary variables.
Ptr< ScalarController< Real > > values_
void getHessVec(Vector< Real > &hv, std::vector< Real > &hvstat, const Vector< Real > &v, const std::vector< Real > &vstat, const Vector< Real > &x, const std::vector< Real > &xstat, SampleGenerator< Real > &sampler)
Return risk measure Hessian-times-a-vector.
Real getValue(const Vector< Real > &x, const std::vector< Real > &xstat, SampleGenerator< Real > &sampler)
Return risk measure value.
Ptr< ScalarController< Real > > gradvecs_
Ptr< VectorController< Real > > hessvecs_
void getGradient(Vector< Real > &g, std::vector< Real > &gstat, const Vector< Real > &x, const std::vector< Real > &xstat, SampleGenerator< Real > &sampler)
Return risk measure (sub)gradient.
void updateValue(Objective< Real > &obj, const Vector< Real > &x, const std::vector< Real > &xstat, Real &tol)
Update internal storage for value computation.
void setHessVecStorage(const Ptr< ScalarController< Real > > &gradvec_storage, const Ptr< VectorController< Real > > &hessvec_storage)
void getMultiplier(Real &lam, const std::vector< Real > &pt) const
Real ppf(const Real x, const Real t, const Real r, const int deriv=0) const
virtual void setHessVecStorage(const Ptr< ScalarController< Real > > &gradvec_storage, const Ptr< VectorController< Real > > &hessvec_storage)
void setValue(const Real val, const std::vector< Real > &pt)
virtual void initialize(const Vector< Real > &x)
Initialize temporary variables.
virtual void setStorage(const Ptr< ScalarController< Real > > &value_storage, const Ptr< VectorController< Real > > &gradient_storage)
Provides the interface to implement any functional that maps a random variable to a (extended) real n...
Real computeValue(Objective< Real > &obj, const Vector< Real > &x, Real &tol)
Ptr< Vector< Real > > g_
void computeHessVec(Vector< Real > &hv, Objective< Real > &obj, const Vector< Real > &v, const Vector< Real > &x, Real &tol)
virtual void setStorage(const Ptr< ScalarController< Real > > &value_storage, const Ptr< VectorController< Real > > &gradient_storage)
Ptr< Vector< Real > > hv_
void computeGradient(Vector< Real > &g, Objective< Real > &obj, const Vector< Real > &x, Real &tol)
Ptr< Vector< Real > > dualVector_
virtual void setHessVecStorage(const Ptr< ScalarController< Real > > &gradvec_storage, const Ptr< VectorController< Real > > &hessvec_storage)
Real computeGradVec(Vector< Real > &g, Objective< Real > &obj, const Vector< Real > &v, const Vector< Real > &x, Real &tol)
virtual int numMySamples(void) const
virtual std::vector< Real > getMyPoint(const int i) const
void sumAll(Real *input, Real *output, int dim) const
virtual Real getMyWeight(const int i) const
Defines the linear algebra or vector space interface.
Definition: ROL_Vector.hpp:84
virtual void plus(const Vector &x)=0
Compute , where .