| 1 |
/* -*- mode: c++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*- */ |
| 2 |
|
| 3 |
/* |
| 4 |
Copyright (C) 2006 Ferdinando Ametrano |
| 5 |
Copyright (C) 2009 Frédéric Degraeve |
| 6 |
|
| 7 |
This file is part of QuantLib, a free-software/open-source library |
| 8 |
for financial quantitative analysts and developers - http://quantlib.org/ |
| 9 |
|
| 10 |
QuantLib is free software: you can redistribute it and/or modify it |
| 11 |
under the terms of the QuantLib license. You should have received a |
| 12 |
copy of the license along with this program; if not, please email |
| 13 |
<quantlib-dev@lists.sf.net>. The license is also available online at |
| 14 |
<http://quantlib.org/license.shtml>. |
| 15 |
|
| 16 |
This program is distributed in the hope that it will be useful, but WITHOUT |
| 17 |
ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS |
| 18 |
FOR A PARTICULAR PURPOSE. See the license for more details. |
| 19 |
*/ |
| 20 |
|
| 21 |
#include "optimization/LineSearchBasedMethod.hpp" |
| 22 |
#include "optimization/Problem.hpp" |
| 23 |
#include "optimization/LineSearch.hpp" |
| 24 |
#include "optimization/Armijo.hpp" |
| 25 |
#include "utils/NumericConstant.hpp" |
| 26 |
|
| 27 |
using namespace OpenMD; |
| 28 |
namespace QuantLib { |
| 29 |
|
| 30 |
LineSearchBasedMethod::LineSearchBasedMethod(LineSearch* lineSearch) |
| 31 |
: lineSearch_(lineSearch) { |
| 32 |
if (!lineSearch_) |
| 33 |
lineSearch_ = new ArmijoLineSearch(); |
| 34 |
} |
| 35 |
|
| 36 |
EndCriteria::Type |
| 37 |
LineSearchBasedMethod::minimize(Problem& P, |
| 38 |
const EndCriteria& endCriteria) { |
| 39 |
// Initializations |
| 40 |
RealType ftol = endCriteria.functionEpsilon(); |
| 41 |
size_t maxStationaryStateIterations_ |
| 42 |
= endCriteria.maxStationaryStateIterations(); |
| 43 |
EndCriteria::Type ecType = EndCriteria::None; // reset end criteria |
| 44 |
P.reset(); // reset problem |
| 45 |
DynamicVector<RealType> x_ = P.currentValue(); // store the starting point |
| 46 |
size_t iterationNumber_ = 0; |
| 47 |
// dimension line search |
| 48 |
lineSearch_->searchDirection() = DynamicVector<RealType>(x_.size()); |
| 49 |
bool done = false; |
| 50 |
|
| 51 |
// function and squared norm of gradient values; |
| 52 |
RealType fnew, fold, gold2; |
| 53 |
RealType fdiff; |
| 54 |
// classical initial value for line-search step |
| 55 |
RealType t = 1.0; |
| 56 |
// Set gradient g at the size of the optimization problem |
| 57 |
// search direction |
| 58 |
size_t sz = lineSearch_->searchDirection().size(); |
| 59 |
DynamicVector<RealType> prevGradient(sz), d(sz), sddiff(sz), direction(sz); |
| 60 |
// Initialize objective function, gradient prevGradient and |
| 61 |
// search direction |
| 62 |
P.setFunctionValue(P.valueAndGradient(prevGradient, x_)); |
| 63 |
P.setGradientNormValue(P.DotProduct(prevGradient, prevGradient)); |
| 64 |
lineSearch_->searchDirection() = -prevGradient; |
| 65 |
|
| 66 |
bool first_time = true; |
| 67 |
// Loop over iterations |
| 68 |
do { |
| 69 |
// Linesearch |
| 70 |
if (!first_time) |
| 71 |
prevGradient = lineSearch_->lastGradient(); |
| 72 |
t = (*lineSearch_)(P, ecType, endCriteria, t); |
| 73 |
// don't throw: it can fail just because maxIterations exceeded |
| 74 |
//QL_REQUIRE(lineSearch_->succeed(), "line-search failed!"); |
| 75 |
if (lineSearch_->succeed()) |
| 76 |
{ |
| 77 |
// Updates |
| 78 |
// New point |
| 79 |
x_ = lineSearch_->lastX(); |
| 80 |
// New function value |
| 81 |
fold = P.functionValue(); |
| 82 |
P.setFunctionValue(lineSearch_->lastFunctionValue()); |
| 83 |
// New gradient and search direction vectors |
| 84 |
|
| 85 |
// orthogonalization coef |
| 86 |
gold2 = P.gradientNormValue(); |
| 87 |
P.setGradientNormValue(lineSearch_->lastGradientNorm2()); |
| 88 |
|
| 89 |
// conjugate gradient search direction |
| 90 |
direction = getUpdatedDirection(P, gold2, prevGradient); |
| 91 |
sddiff = direction - lineSearch_->searchDirection(); |
| 92 |
lineSearch_->searchDirection() = direction; |
| 93 |
// Now compute accuracy and check end criteria |
| 94 |
// Numerical Recipes exit strategy on fx (see NR in C++, p.423) |
| 95 |
|
| 96 |
fnew = P.functionValue(); |
| 97 |
fdiff = 2.0*std::fabs(fnew-fold) / |
| 98 |
(std::fabs(fnew) + std::fabs(fold) + NumericConstant::epsilon); |
| 99 |
|
| 100 |
if (fdiff < ftol || |
| 101 |
endCriteria.checkMaxIterations(iterationNumber_, ecType)) { |
| 102 |
endCriteria.checkStationaryFunctionValue(0.0, 0.0, |
| 103 |
maxStationaryStateIterations_, ecType); |
| 104 |
endCriteria.checkMaxIterations(iterationNumber_, ecType); |
| 105 |
return ecType; |
| 106 |
} |
| 107 |
|
| 108 |
P.setCurrentValue(x_); // update problem current value |
| 109 |
++iterationNumber_; // Increase iteration number |
| 110 |
first_time = false; |
| 111 |
} else { |
| 112 |
done = true; |
| 113 |
} |
| 114 |
} while (!done); |
| 115 |
P.setCurrentValue(x_); |
| 116 |
return ecType; |
| 117 |
} |
| 118 |
|
| 119 |
} |
