MoochoPack_DampenCrossTermStd_Step.cpp

#if 0

// @HEADER
// ***********************************************************************
// 
// Moocho: Multi-functional Object-Oriented arCHitecture for Optimization
//                  Copyright (2003) 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.
// 
// This library is free software; you can redistribute it and/or modify
// it under the terms of the GNU Lesser General Public License as
// published by the Free Software Foundation; either version 2.1 of the
// License, or (at your option) any later version.
//  
// This library is distributed in the hope that it will be useful, but
// WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
// Lesser General Public License for more details.
//  
// You should have received a copy of the GNU Lesser General Public
// License along with this library; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307
// USA
// Questions? Contact Roscoe A. Bartlett (rabartl@sandia.gov) 
// 
// ***********************************************************************
// @HEADER

#include <ostream>
#include <sstream>
#include <limits>

#include "MoochoPack_DampenCrossTermStd_Step.hpp"
#include "MoochoPack_moocho_algo_conversion.hpp"
#include "IterationPack_print_algorithm_step.hpp"
#include "ConstrainedOptPack/src/VectorWithNorms.h"
#include "AbstractLinAlgPack/src/MatrixWithOpFactorized.hpp"
#include "DenseLinAlgPack_LinAlgOpPack.hpp"
#include "DenseLinAlgPack_DVectorClass.hpp"
#include "DenseLinAlgPack_DVectorOut.hpp"

MoochoPack::DampenCrossTermStd_Step::DampenCrossTermStd_Step(const value_type& frac_descent)
  : frac_descent_(frac_descent)
{}

bool MoochoPack::DampenCrossTermStd_Step::do_step(Algorithm& _algo
  , poss_type step_poss, IterationPack::EDoStepType type, poss_type assoc_step_poss)
{
  using AbstractLinAlgPack::V_InvMtV;
  using DenseLinAlgPack::norm_inf;
  using DenseLinAlgPack::dot;

  NLPAlgo &algo = rsqp_algo(_algo);
  NLPAlgoState  &s    = algo.rsqp_state();

  EJournalOutputLevel olevel = algo.algo_cntr().journal_output_level();
  std::ostream& out = algo.track().journal_out();

  // print step header.
  if( static_cast<int>(olevel) >= static_cast<int>(PRINT_ALGORITHM_STEPS) ) {
    using IterationPack::print_algorithm_step;
    print_algorithm_step( _algo, step_poss, type, assoc_step_poss, out );
  }

  if( s.w().updated_k(0) ) {

    // inv(rHL_k) * rGf_k
    const DVectorSlice rGf_k = s.rGf().get_k(0)();
    DVector Inv_rHL_rGf;
    V_InvMtV( &Inv_rHL_rGf, dynamic_cast<MatrixWithOpFactorized&>(s.rHL().get_k(0))
      , BLAS_Cpp::no_trans, rGf_k );
    
    const value_type
      small_num     = 1e-20,
      rGfT_Inv_rHL_rGf  = dot( Inv_rHL_rGf(), rGf_k ),          // rGf_k'*inv(rHL_k)*rGf_k
      rGfT_Inv_rHL_w    = dot( Inv_rHL_rGf(), s.w().get_k(0)() ),   // rGf_k'*inv(rHL_k)*w_k
      term        = -(1.0-frac_descent()) * (rGfT_Inv_rHL_rGf + 2*small_num)
                  / (rGfT_Inv_rHL_w + small_num);

    if( rGfT_Inv_rHL_w >= 0.0 ) {
      // We know that the descent property will be satisfied for all zeta_k > 0
      // so set zeta_k = 1
      s.zeta().set_k(0) = 1.0;
    }
    else {
      // For some zeta_k > 0 the descent property will be violated so we may have to
      // cut zeta_k back from 1.
      s.zeta().set_k(0) = std::_MIN( term, 1.0 );
    }

    if( static_cast<int>(olevel) >= static_cast<int>(PRINT_ALGORITHM_STEPS) ) {
      out << "\nterm1 = rGf_k'*inv(rHL_k)*rGf_k                = "  <<  rGfT_Inv_rHL_rGf;
      out << "\nterm2 = rGf_k'*inv(rHL_k)*w_k                  = "  <<  rGfT_Inv_rHL_w;
      out << "\n(1-frac_descent)*(term1+2*small)/(term2+small) = "  <<  term;
      out << "\nzeta_k                                         = "  <<  s.zeta().get_k(0)
        << std::endl;
    }

    if( static_cast<int>(olevel) >= static_cast<int>(PRINT_VECTORS) ) {
      out << "\ninv(rHL_k)*rGf_k = "  <<  Inv_rHL_rGf();
    }

    if( rGfT_Inv_rHL_rGf < 0.0 ) {
      std::ostringstream omsg;
      omsg
        << "Error, rGf_k'*inv(rHL_k)*rGf_k = " << rGfT_Inv_rHL_rGf << " < 0.0 and therefore "
        << "the reduced Hessian rHL_k can not be positive definite";
      if( (int)(olevel) >= (int)(PRINT_ALGORITHM_STEPS) ) {
        out << omsg.str();
      }
      throw std::runtime_error( std::string("DampenCrossTermStd_Step::do_step(...) : ")
                    + omsg.str() );
    }
  }

  return true;
}

void MoochoPack::DampenCrossTermStd_Step::print_step( const Algorithm& algo
  , poss_type step_poss, IterationPack::EDoStepType type, poss_type assoc_step_poss
  , std::ostream& out, const std::string& L ) const
{
  out
    << L << "*** Compute the dampening parameter for the reduced QP cross term w_k\n"
    << L << "default: frac_descent = " << frac_descent() << std::endl
    << L << "if w_k is update then\n"
    << L << "  find zeta_k s.t.\n"
    << L << "    Gf_k'*Z_k*pz_k approx\n"
    << L << "       - zeta_k * rGf_k'*inv(rHL_k)*w_k - rGf_k'*inv(rHL_k)*rGf_k\n"
    << L << "       <= - frac_descent * rGf_k'*inv(rHL_k)*rGf_k\n"
    << L << "end\n";
}

#endif // 0