| 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 <ql/math/optimization/armijo.hpp> |
| 22 | #include <ql/math/optimization/linesearch.hpp> |
| 23 | #include <ql/math/optimization/linesearchbasedmethod.hpp> |
| 24 | #include <ql/math/optimization/problem.hpp> |
| 25 | #include <utility> |
| 26 | |
| 27 | namespace QuantLib { |
| 28 | |
| 29 | LineSearchBasedMethod::LineSearchBasedMethod(ext::shared_ptr<LineSearch> lineSearch) |
| 30 | : lineSearch_(std::move(lineSearch)) { |
| 31 | if (!lineSearch_) |
| 32 | lineSearch_ = ext::shared_ptr<LineSearch>(new ArmijoLineSearch); |
| 33 | } |
| 34 | |
| 35 | EndCriteria::Type |
| 36 | LineSearchBasedMethod::minimize(Problem& P, |
| 37 | const EndCriteria& endCriteria) { |
| 38 | // Initializations |
| 39 | Real ftol = endCriteria.functionEpsilon(); |
| 40 | Size maxStationaryStateIterations_ |
| 41 | = endCriteria.maxStationaryStateIterations(); |
| 42 | EndCriteria::Type ecType = EndCriteria::None; // reset end criteria |
| 43 | P.reset(); // reset problem |
| 44 | Array x_ = P.currentValue(); // store the starting point |
| 45 | Size iterationNumber_ = 0; |
| 46 | // dimension line search |
| 47 | lineSearch_->searchDirection() = Array(x_.size()); |
| 48 | bool done = false; |
| 49 | |
| 50 | // function and squared norm of gradient values; |
| 51 | Real fnew, fold, gold2; |
| 52 | Real fdiff; |
| 53 | // classical initial value for line-search step |
| 54 | Real t = 1.0; |
| 55 | // Set gradient g at the size of the optimization problem |
| 56 | // search direction |
| 57 | Size sz = lineSearch_->searchDirection().size(); |
| 58 | Array prevGradient(sz), d(sz), sddiff(sz), direction(sz); |
| 59 | // Initialize cost function, gradient prevGradient and search direction |
| 60 | P.setFunctionValue(P.valueAndGradient(grad_f&: prevGradient, x: x_)); |
| 61 | P.setGradientNormValue(DotProduct(v1: prevGradient, v2: prevGradient)); |
| 62 | lineSearch_->searchDirection() = -prevGradient; |
| 63 | |
| 64 | bool first_time = true; |
| 65 | // Loop over iterations |
| 66 | do { |
| 67 | // Linesearch |
| 68 | if (!first_time) |
| 69 | prevGradient = lineSearch_->lastGradient(); |
| 70 | t = (*lineSearch_)(P, ecType, endCriteria, t); |
| 71 | // don't throw: it can fail just because maxIterations exceeded |
| 72 | //QL_REQUIRE(lineSearch_->succeed(), "line-search failed!"); |
| 73 | if (lineSearch_->succeed()) |
| 74 | { |
| 75 | // Updates |
| 76 | |
| 77 | // New point |
| 78 | x_ = lineSearch_->lastX(); |
| 79 | // New function value |
| 80 | fold = P.functionValue(); |
| 81 | P.setFunctionValue(lineSearch_->lastFunctionValue()); |
| 82 | // New gradient and search direction vectors |
| 83 | |
| 84 | // orthogonalization coef |
| 85 | gold2 = P.gradientNormValue(); |
| 86 | P.setGradientNormValue(lineSearch_->lastGradientNorm2()); |
| 87 | |
| 88 | // conjugate gradient search direction |
| 89 | direction = getUpdatedDirection(P, gold2, gradient: prevGradient); |
| 90 | |
| 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 | fnew = P.functionValue(); |
| 96 | fdiff = 2.0*std::fabs(x: fnew-fold) / |
| 97 | (std::fabs(x: fnew) + std::fabs(x: fold) + QL_EPSILON); |
| 98 | if (fdiff < ftol || |
| 99 | endCriteria.checkMaxIterations(iteration: iterationNumber_, ecType)) { |
| 100 | endCriteria.checkStationaryFunctionValue(fxOld: 0.0, fxNew: 0.0, |
| 101 | statStateIterations&: maxStationaryStateIterations_, ecType); |
| 102 | endCriteria.checkMaxIterations(iteration: iterationNumber_, ecType); |
| 103 | return ecType; |
| 104 | } |
| 105 | P.setCurrentValue(x_); // update problem current value |
| 106 | ++iterationNumber_; // Increase iteration number |
| 107 | first_time = false; |
| 108 | } else { |
| 109 | done = true; |
| 110 | } |
| 111 | } while (!done); |
| 112 | P.setCurrentValue(x_); |
| 113 | return ecType; |
| 114 | } |
| 115 | |
| 116 | } |
| 117 | |