ROL_FDivergence.hpp

Go to the documentation of this file.

// @HEADER
// ************************************************************************
//
//               Rapid Optimization Library (ROL) Package
//                 Copyright (2014) Sandia Corporation
//
// Under terms of Contract DE-AC04-94AL85000, there is a non-exclusive
// license for use of this work by or on behalf of the U.S. Government.
//
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are
// met:
//
// 1. Redistributions of source code must retain the above copyright
// notice, this list of conditions and the following disclaimer.
//
// 2. Redistributions in binary form must reproduce the above copyright
// notice, this list of conditions and the following disclaimer in the
// documentation and/or other materials provided with the distribution.
//
// 3. Neither the name of the Corporation nor the names of the
// contributors may be used to endorse or promote products derived from
// this software without specific prior written permission.
//
// THIS SOFTWARE IS PROVIDED BY SANDIA CORPORATION "AS IS" AND ANY
// EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
// IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
// PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL SANDIA CORPORATION OR THE
// CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
// EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
// PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
// PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
// LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
// NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
// SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
//
// Questions? Contact lead developers:
//              Drew Kouri   (dpkouri@sandia.gov) and
//              Denis Ridzal (dridzal@sandia.gov)
//
// ************************************************************************
// @HEADER

#ifndef ROL_FDIVERGENCE_HPP
#define ROL_FDIVERGENCE_HPP

#include "ROL_RandVarFunctional.hpp"
#include "ROL_Types.hpp"

namespace ROL {

template<class Real>
class FDivergence : public RandVarFunctional<Real> {
private:
  Real thresh_;

  Real valLam_;
  Real valLam2_;
  Real valMu_;
  Real valMu2_;

  using RandVarFunctional<Real>::val_;
  using RandVarFunctional<Real>::gv_;
  using RandVarFunctional<Real>::g_;
  using RandVarFunctional<Real>::hv_;
  using RandVarFunctional<Real>::dualVector_;

  using RandVarFunctional<Real>::point_;
  using RandVarFunctional<Real>::weight_;

  using RandVarFunctional<Real>::computeValue;
  using RandVarFunctional<Real>::computeGradient;
  using RandVarFunctional<Real>::computeGradVec;
  using RandVarFunctional<Real>::computeHessVec;

  void checkInputs(void) const {
    Real zero(0);
    ROL_TEST_FOR_EXCEPTION((thresh_ <= zero), std::invalid_argument,
      ">>> ERROR (ROL::FDivergence): Threshold must be positive!");
  }

public:
  FDivergence(const Real thresh) : RandVarFunctional<Real>(), thresh_(thresh),
    valLam_(0),valLam2_(0), valMu_(0), valMu2_(0) {
    checkInputs();
  }

  FDivergence(ROL::ParameterList &parlist) : RandVarFunctional<Real>(),
    valLam_(0),valLam2_(0), valMu_(0), valMu2_(0) {
    ROL::ParameterList &list
      = parlist.sublist("SOL").sublist("Risk Measure").sublist("F-Divergence");
    thresh_ = list.get<Real>("Threshold");
    checkInputs();
  }

  virtual Real Fprimal(Real x, int deriv = 0) = 0;

  virtual Real Fdual(Real x, int deriv = 0) = 0;

  bool check(std::ostream &outStream = std::cout) const {
    const Real tol(std::sqrt(ROL_EPSILON<Real>()));
    bool flag = true;

    Real x  = static_cast<Real>(rand())/static_cast<Real>(RAND_MAX);
    Real t  = static_cast<Real>(rand())/static_cast<Real>(RAND_MAX);
    Real fp = Fprimal(x);
    Real fd = Fdual(t);
    outStream << "Check Fenchel-Young Inequality: F(x) + F*(t) >= xt" << std::endl;
    outStream << "x       = " << x                << std::endl;
    outStream << "t       = " << t                << std::endl;
    outStream << "F(x)    = " << fp               << std::endl;
    outStream << "F*(t)   = " << fd               << std::endl;
    outStream << "Is Valid? " << (fp+fd >= x*t)   << std::endl;
    flag = (fp+fd >= x*t) ? flag : false;

    x  = static_cast<Real>(rand())/static_cast<Real>(RAND_MAX);
    t  = Fprimal(x,1);
    fp = Fprimal(x);
    fd = Fdual(t);
    outStream << "Check Fenchel-Young Equality: F(x) + F(t) = xt for t = d/dx F(x)" << std::endl;
    outStream << "x       = " << x                << std::endl;
    outStream << "t       = " << t                << std::endl;
    outStream << "F(x)    = " << fp               << std::endl;
    outStream << "F*(t)   = " << fd               << std::endl;
    outStream << "Is Valid? " << (std::abs(fp+fd - x*t)<=tol) << std::endl; 
    flag = (std::abs(fp+fd - x*t)<=tol) ? flag : false;

    t  = static_cast<Real>(rand())/static_cast<Real>(RAND_MAX);
    x  = Fdual(t,1);
    fp = Fprimal(x);
    fd = Fdual(t);
    outStream << "Check Fenchel-Young Equality: F(x) + F(t) = xt for x = d/dt F*(t)" << std::endl;
    outStream << "x       = " << x                << std::endl;
    outStream << "t       = " << t                << std::endl;
    outStream << "F(x)    = " << fp               << std::endl;
    outStream << "F*(t)   = " << fd               << std::endl;
    outStream << "Is Valid? " << (std::abs(fp+fd - x*t)<=tol) << std::endl; 
    flag = (std::abs(fp+fd - x*t)<=tol) ? flag : false;

    return flag;
  }

  void initialize(const Vector<Real> &x) {
    RandVarFunctional<Real>::initialize(x);
    valLam_ = 0; valLam2_ = 0; valMu_ = 0; valMu2_ = 0;
  }

  // Value update and get functions
  void updateValue(Objective<Real>         &obj,
                   const Vector<Real>      &x,
                   const std::vector<Real> &xstat,
                   Real                    &tol) {
    Real val  = computeValue(obj,x,tol);
    Real xlam = xstat[0];
    Real xmu  = xstat[1];
    Real r    = Fdual((val-xmu)/xlam,0);
    val_     += weight_ * r;
  }

  Real getValue(const Vector<Real>      &x,
                const std::vector<Real> &xstat,
                SampleGenerator<Real>   &sampler) {
    Real val(0);
    sampler.sumAll(&val_,&val,1);
    Real xlam = xstat[0];
    Real xmu  = xstat[1];
    return xlam*(thresh_ + val) + xmu;
  }

  // Gradient update and get functions
  void updateGradient(Objective<Real>         &obj,
                      const Vector<Real>      &x,
                      const std::vector<Real> &xstat,
                      Real                    &tol) {
    Real val  = computeValue(obj,x,tol);
    Real xlam = xstat[0];
    Real xmu  = xstat[1];
    Real inp  = (val-xmu)/xlam;
    Real r0 = Fdual(inp,0), r1 = Fdual(inp,1);

    if (std::abs(r0) >= ROL_EPSILON<Real>()) {
      val_ += weight_ * r0;
    }
    if (std::abs(r1) >= ROL_EPSILON<Real>()) {
      valLam_ -= weight_ * r1 * inp;
      valMu_  -= weight_ * r1;
      computeGradient(*dualVector_,obj,x,tol);
      g_->axpy(weight_*r1,*dualVector_);
    }
  }

  void getGradient(Vector<Real>            &g,
                   std::vector<Real>       &gstat,
                   const Vector<Real>      &x,
                   const std::vector<Real> &xstat,
                   SampleGenerator<Real>   &sampler) {
    std::vector<Real> mygval(3), gval(3);
    mygval[0] = val_;
    mygval[1] = valLam_;
    mygval[2] = valMu_;
    sampler.sumAll(&mygval[0],&gval[0],3);

    gstat[0] = thresh_ + gval[0] + gval[1];
    gstat[1] = static_cast<Real>(1) + gval[2];

    sampler.sumAll(*g_,g);
  }

  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) {
    Real val  = computeValue(obj,x,tol);
    Real xlam = xstat[0];
    Real xmu  = xstat[1];
    Real vlam = vstat[0];
    Real vmu  = vstat[1];
    Real inp  = (val-xmu)/xlam;
    Real r1 = Fdual(inp,1), r2 = Fdual(inp,2);
    if (std::abs(r2) >= ROL_EPSILON<Real>()) {
      Real gv   = computeGradVec(*dualVector_,obj,v,x,tol);
      val_     += weight_ * r2 * inp;
      valLam_  += weight_ * r2 * inp * inp;
      valLam2_ -= weight_ * r2 * gv * inp;
      valMu_   += weight_ * r2;
      valMu2_  -= weight_ * r2 * gv;
      hv_->axpy(weight_ * r2 * (gv - vmu - vlam*inp)/xlam, *dualVector_);
    }
    if (std::abs(r1) >= ROL_EPSILON<Real>()) {
      computeHessVec(*dualVector_,obj,v,x,tol);
      hv_->axpy(weight_ * r1, *dualVector_);
    }
  }

  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) {
    std::vector<Real> myhval(5), hval(5);
    myhval[0] = val_;
    myhval[1] = valLam_;
    myhval[2] = valLam2_;
    myhval[3] = valMu_;
    myhval[4] = valMu2_;
    sampler.sumAll(&myhval[0],&hval[0],5);

    std::vector<Real> stat(2);
    Real xlam = xstat[0];
    //Real xmu  = xstat[1];
    Real vlam = vstat[0];
    Real vmu  = vstat[1];
    hvstat[0] = (vlam * hval[1] + vmu * hval[0] + hval[2])/xlam;
    hvstat[1] = (vlam * hval[0] + vmu * hval[3] + hval[4])/xlam;

    sampler.sumAll(*hv_,hv);
  }
};

}

#endif