27603489947

Committed 16 Jun 2026 08:07AM UTC coverage: 74.828% (-0.006%) from 74.834%

Build # 27603489947

Build Type

Pull #2629

github

Committed by

web-flow

Commit Message

Merge a100f20fd into 9eac92aff

Pull Request Pull Request #2629: Replace Boost macros that might not work with AAD types

Coverage Stats

59233 of 79159 relevant lines covered (74.83%)

8636679.49 hits per line

Source File
Press 'n' to go to next uncovered line, 'b' for previous

84.21

/ql/termstructures/volatility/abcdcalibration.cpp

/* -*- mode: c++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*- */

/*
 Copyright (C) 2006, 2015 Ferdinando Ametrano
 Copyright (C) 2006 Cristina Duminuco
 Copyright (C) 2005, 2006 Klaus Spanderen
 Copyright (C) 2007 Giorgio Facchinetti
 Copyright (C) 2015 Paolo Mazzocchi

 This file is part of QuantLib, a free-software/open-source library
 for financial quantitative analysts and developers - http://quantlib.org/

 QuantLib is free software: you can redistribute it and/or modify it
 under the terms of the QuantLib license.  You should have received a
 copy of the license along with this program; if not, please email
 <quantlib-dev@lists.sf.net>. The license is also available online at
 <https://www.quantlib.org/license.shtml>.

 This program 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 license for more details.
*/

#include <ql/math/distributions/normaldistribution.hpp>
#include <ql/math/interpolations/abcdinterpolation.hpp>
#include <ql/math/optimization/constraint.hpp>
#include <ql/math/optimization/levenbergmarquardt.hpp>
#include <ql/math/optimization/method.hpp>
#include <ql/pricingengines/blackformula.hpp>
#include <ql/termstructures/volatility/abcd.hpp>
#include <ql/termstructures/volatility/abcdcalibration.hpp>
#include <utility>

namespace QuantLib {

    // to constrained <- from unconstrained
    Array AbcdCalibration::AbcdParametersTransformation::direct(const Array& x) const {
        y_[1] = x[1];
        y_[2] = std::exp(x[2]);
        y_[3] = std::exp(x[3]);
        y_[0] = std::exp(x[0]) - y_[3];
        return y_;
    }

    // to unconstrained <- from constrained
    Array AbcdCalibration::AbcdParametersTransformation::inverse(const Array& x) const {
        y_[1] = x[1];
        y_[2] = std::log(x[2]);
        y_[3] = std::log(x[3]);
        y_[0] = std::log(x[0] + x[3]);
        return y_;
    }

    // to constrained <- from unconstrained

    AbcdCalibration::AbcdCalibration(const std::vector<Real>& t,
                                     const std::vector<Real>& blackVols,
                                     Real a,
                                     Real b,
                                     Real c,
                                     Real d,
                                     bool aIsFixed,
                                     bool bIsFixed,
                                     bool cIsFixed,
                                     bool dIsFixed,
                                     bool vegaWeighted,
                                     ext::shared_ptr<EndCriteria> endCriteria,
                                     ext::shared_ptr<OptimizationMethod> optMethod)
    : aIsFixed_(aIsFixed), bIsFixed_(bIsFixed), cIsFixed_(cIsFixed), dIsFixed_(dIsFixed), a_(a),
      b_(b), c_(c), d_(d), abcdEndCriteria_(EndCriteria::None),
      endCriteria_(std::move(endCriteria)), optMethod_(std::move(optMethod)),
      weights_(blackVols.size(), 1.0 / blackVols.size()), vegaWeighted_(vegaWeighted), times_(t),
      blackVols_(blackVols) {

        AbcdMathFunction::validate(a, b, c, d);

        QL_REQUIRE(blackVols.size()==t.size(),
                       "mismatch between number of times (" << t.size() <<
                       ") and blackVols (" << blackVols.size() << ")");

        // if no optimization method or endCriteria is provided, we provide one
        if (!optMethod_) {
            Real epsfcn = 1.0e-8;
            Real xtol = 1.0e-8;
            Real gtol = 1.0e-8;
            bool useCostFunctionsJacobian = false;
            optMethod_ = ext::shared_ptr<OptimizationMethod>(new
                LevenbergMarquardt(epsfcn, xtol, gtol, useCostFunctionsJacobian));
        }
        if (!endCriteria_) {
            Size maxIterations = 10000;
            Size maxStationaryStateIterations = 1000;
            Real rootEpsilon = 1.0e-8;
            Real functionEpsilon = 0.3e-4;     // Why 0.3e-4 ?
            Real gradientNormEpsilon = 0.3e-4; // Why 0.3e-4 ?
            endCriteria_ = ext::make_shared<EndCriteria>(maxIterations, maxStationaryStateIterations,
                            rootEpsilon, functionEpsilon, gradientNormEpsilon);
        }
    }

    void AbcdCalibration::compute() {
        if (vegaWeighted_) {
            Real weightsSum = 0.0;
            for (Size i=0; i<times_.size() ; i++) {
                Real stdDev = std::sqrt(blackVols_[i]* blackVols_[i]* times_[i]);
                // when strike==forward, the blackFormulaStdDevDerivative becomes
                weights_[i] = CumulativeNormalDistribution().derivative(.5*stdDev);
                weightsSum += weights_[i];
            }
            // weight normalization
            for (Size i=0; i<times_.size() ; i++) {
                weights_[i] /= weightsSum;
            }
        }

        // there is nothing to optimize
        if (aIsFixed_ && bIsFixed_ && cIsFixed_ && dIsFixed_) {
            abcdEndCriteria_ = EndCriteria::None;
            //error_ = interpolationError();
            //maxError_ = interpolationMaxError();
            return;
        } else {

            AbcdError costFunction(this);
            transformation_ = ext::shared_ptr<ParametersTransformation>(new
                AbcdParametersTransformation);

            Array guess(4);
            guess[0] = a_;
            guess[1] = b_;
            guess[2] = c_;
            guess[3] = d_;

            std::vector<bool> parameterAreFixed(4);
            parameterAreFixed[0] = aIsFixed_;
            parameterAreFixed[1] = bIsFixed_;
            parameterAreFixed[2] = cIsFixed_;
            parameterAreFixed[3] = dIsFixed_;

            Array inversedTransformatedGuess(transformation_->inverse(guess));

            ProjectedCostFunction projectedAbcdCostFunction(costFunction,
                            inversedTransformatedGuess, parameterAreFixed);

            Array projectedGuess
                (projectedAbcdCostFunction.project(inversedTransformatedGuess));

            NoConstraint constraint;
            Problem problem(projectedAbcdCostFunction, constraint, projectedGuess);
            abcdEndCriteria_ = optMethod_->minimize(problem, *endCriteria_);
            Array projectedResult(problem.currentValue());
            Array transfResult(projectedAbcdCostFunction.include(projectedResult));

            Array result = transformation_->direct(transfResult);
            AbcdMathFunction::validate(a_, b_, c_, d_);
            a_ = result[0];
            b_ = result[1];
            c_ = result[2];
            d_ = result[3];

        }
    }

    Real AbcdCalibration::value(Real x) const {
        return abcdBlackVolatility(x,a_,b_,c_,d_);
    }

    std::vector<Real> AbcdCalibration::k(const std::vector<Real>& t,
                                         const std::vector<Real>& blackVols) const {
        QL_REQUIRE(blackVols.size()==t.size(),
               "mismatch between number of times (" << t.size() <<
               ") and blackVols (" << blackVols.size() << ")");
        std::vector<Real> k(t.size());
        for (Size i=0; i<t.size() ; i++) {
            k[i]=blackVols[i]/value(t[i]);
        }
        return k;
    }

    Real AbcdCalibration::error() const {
        Size n = times_.size();
        Real error, squaredError = 0.0;
        for (Size i=0; i<times_.size() ; i++) {
            error = (value(times_[i]) - blackVols_[i]);
            squaredError += error * error * weights_[i];
        }
        return std::sqrt(n*squaredError/(n-1));
    }

    Real AbcdCalibration::maxError() const {
        Real error, maxError = QL_MIN_REAL;
        for (Size i=0; i<times_.size() ; i++) {
            error = std::fabs(value(times_[i]) - blackVols_[i]);
            maxError = std::max(maxError, error);
        }
        return maxError;
    }

    // calculate weighted differences
    Array AbcdCalibration::errors() const {
        Array results(times_.size());
        for (Size i=0; i<times_.size() ; i++) {
            results[i] = (value(times_[i]) - blackVols_[i])* std::sqrt(weights_[i]);
        }
        return results;
    }

    EndCriteria::Type AbcdCalibration::endCriteria() const{
        return abcdEndCriteria_;
    }

}

1	/* -- mode: c++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -- */
2
3	/*
4	Copyright (C) 2006, 2015 Ferdinando Ametrano
5	Copyright (C) 2006 Cristina Duminuco
6	Copyright (C) 2005, 2006 Klaus Spanderen
7	Copyright (C) 2007 Giorgio Facchinetti
8	Copyright (C) 2015 Paolo Mazzocchi
9
10	This file is part of QuantLib, a free-software/open-source library
11	for financial quantitative analysts and developers - http://quantlib.org/
12
13	QuantLib is free software: you can redistribute it and/or modify it
14	under the terms of the QuantLib license. You should have received a
15	copy of the license along with this program; if not, please email
16	<quantlib-dev@lists.sf.net>. The license is also available online at
17	<https://www.quantlib.org/license.shtml>.
18
19	This program is distributed in the hope that it will be useful, but WITHOUT
20	ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
21	FOR A PARTICULAR PURPOSE. See the license for more details.
22	*/
23
24	#include <ql/math/distributions/normaldistribution.hpp>
25	#include <ql/math/interpolations/abcdinterpolation.hpp>
26	#include <ql/math/optimization/constraint.hpp>
27	#include <ql/math/optimization/levenbergmarquardt.hpp>
28	#include <ql/math/optimization/method.hpp>
29	#include <ql/pricingengines/blackformula.hpp>
30	#include <ql/termstructures/volatility/abcd.hpp>
31	#include <ql/termstructures/volatility/abcdcalibration.hpp>
32	#include <utility>
33
34	namespace QuantLib {
35
36	// to constrained <- from unconstrained
37	Array AbcdCalibration::AbcdParametersTransformation::direct(const Array& x) const {	19✔
38	y_[1] = x[1];	19✔
39	y_[2] = std::exp(x[2]);	19✔
40	y_[3] = std::exp(x[3]);	19✔
41	y_[0] = std::exp(x[0]) - y_[3];	19✔
42	return y_;	19✔
43	}
44
45	// to unconstrained <- from constrained
46	Array AbcdCalibration::AbcdParametersTransformation::inverse(const Array& x) const {	1✔
47	y_[1] = x[1];	1✔
48	y_[2] = std::log(x[2]);	1✔
49	y_[3] = std::log(x[3]);	1✔
50	y_[0] = std::log(x[0] + x[3]);	1✔
51	return y_;	1✔
52	}
53
54	// to constrained <- from unconstrained
55
56	AbcdCalibration::AbcdCalibration(const std::vector<Real>& t,	1✔
57	const std::vector<Real>& blackVols,
58	Real a,
59	Real b,
60	Real c,
61	Real d,
62	bool aIsFixed,
63	bool bIsFixed,
64	bool cIsFixed,
65	bool dIsFixed,
66	bool vegaWeighted,
67	ext::shared_ptr<EndCriteria> endCriteria,
68	ext::shared_ptr<OptimizationMethod> optMethod)	1✔
69	: aIsFixed_(aIsFixed), bIsFixed_(bIsFixed), cIsFixed_(cIsFixed), dIsFixed_(dIsFixed), a_(a),	1✔
70	b_(b), c_(c), d_(d), abcdEndCriteria_(EndCriteria::None),	1✔
71	endCriteria_(std::move(endCriteria)), optMethod_(std::move(optMethod)),	1✔
72	weights_(blackVols.size(), 1.0 / blackVols.size()), vegaWeighted_(vegaWeighted), times_(t),	1✔
73	blackVols_(blackVols) {	1✔
74
75	AbcdMathFunction::validate(a, b, c, d);	1✔
76
77	QL_REQUIRE(blackVols.size()==t.size(),	1✔
78	"mismatch between number of times (" << t.size() <<
79	") and blackVols (" << blackVols.size() << ")");
80
81	// if no optimization method or endCriteria is provided, we provide one
82	if (!optMethod_) {	1✔
83	Real epsfcn = 1.0e-8;
84	Real xtol = 1.0e-8;
85	Real gtol = 1.0e-8;
86	bool useCostFunctionsJacobian = false;
87	optMethod_ = ext::shared_ptr<OptimizationMethod>(new	2✔
88	LevenbergMarquardt(epsfcn, xtol, gtol, useCostFunctionsJacobian));	2✔
89	}
90	if (!endCriteria_) {	1✔
91	Size maxIterations = 10000;	1✔
92	Size maxStationaryStateIterations = 1000;	1✔
93	Real rootEpsilon = 1.0e-8;	1✔
94	Real functionEpsilon = 0.3e-4; // Why 0.3e-4 ?	1✔
95	Real gradientNormEpsilon = 0.3e-4; // Why 0.3e-4 ?	1✔
96	endCriteria_ = ext::make_shared<EndCriteria>(maxIterations, maxStationaryStateIterations,	2✔
97	rootEpsilon, functionEpsilon, gradientNormEpsilon);	1✔
98	}
99	}	1✔
100
101	void AbcdCalibration::compute() {	1✔
102	if (vegaWeighted_) {	1✔
103	Real weightsSum = 0.0;
104	for (Size i=0; i<times_.size() ; i++) {	×
105	Real stdDev = std::sqrt(blackVols_[i]* blackVols_[i]* times_[i]);	×
106	// when strike==forward, the blackFormulaStdDevDerivative becomes
107	weights_[i] = CumulativeNormalDistribution().derivative(.5*stdDev);	×
108	weightsSum += weights_[i];	×
109	}
110	// weight normalization
111	for (Size i=0; i<times_.size() ; i++) {	×
112	weights_[i] /= weightsSum;	×
113	}
114	}
115
116	// there is nothing to optimize
117	if (aIsFixed_ && bIsFixed_ && cIsFixed_ && dIsFixed_) {	1✔
118	abcdEndCriteria_ = EndCriteria::None;	×
119	//error_ = interpolationError();
120	//maxError_ = interpolationMaxError();
121	return;	×
122	} else {
123
124	AbcdError costFunction(this);
125	transformation_ = ext::shared_ptr<ParametersTransformation>(new	2✔
126	AbcdParametersTransformation);	2✔
127
128	Array guess(4);
129	guess[0] = a_;	1✔
130	guess[1] = b_;	1✔
131	guess[2] = c_;	1✔
132	guess[3] = d_;	1✔
133
134	std::vector<bool> parameterAreFixed(4);	×
135	parameterAreFixed[0] = aIsFixed_;	1✔
136	parameterAreFixed[1] = bIsFixed_;	1✔
137	parameterAreFixed[2] = cIsFixed_;	1✔
138	parameterAreFixed[3] = dIsFixed_;	1✔
139
140	Array inversedTransformatedGuess(transformation_->inverse(guess));	1✔
141
142	ProjectedCostFunction projectedAbcdCostFunction(costFunction,
143	inversedTransformatedGuess, parameterAreFixed);	1✔
144
145	Array projectedGuess
146	(projectedAbcdCostFunction.project(inversedTransformatedGuess));	1✔
147
148	NoConstraint constraint;	1✔
149	Problem problem(projectedAbcdCostFunction, constraint, projectedGuess);	1✔
150	abcdEndCriteria_ = optMethod_->minimize(problem, *endCriteria_);	1✔
151	Array projectedResult(problem.currentValue());	1✔
152	Array transfResult(projectedAbcdCostFunction.include(projectedResult));	1✔
153
154	Array result = transformation_->direct(transfResult);	1✔
155	AbcdMathFunction::validate(a_, b_, c_, d_);	1✔
156	a_ = result[0];	1✔
157	b_ = result[1];	1✔
158	c_ = result[2];	1✔
159	d_ = result[3];	1✔
160
161	}
162	}
163
164	Real AbcdCalibration::value(Real x) const {	189✔
165	return abcdBlackVolatility(x,a_,b_,c_,d_);	189✔
166	}
167
168	std::vector<Real> AbcdCalibration::k(const std::vector<Real>& t,	1✔
169	const std::vector<Real>& blackVols) const {
170	QL_REQUIRE(blackVols.size()==t.size(),	1✔
171	"mismatch between number of times (" << t.size() <<
172	") and blackVols (" << blackVols.size() << ")");
173	std::vector<Real> k(t.size());	1✔
174	for (Size i=0; i<t.size() ; i++) {	10✔
175	k[i]=blackVols[i]/value(t[i]);	9✔
176	}
177	return k;	1✔
178	}
179
180	Real AbcdCalibration::error() const {	3✔
181	Size n = times_.size();
182	Real error, squaredError = 0.0;
183	for (Size i=0; i<times_.size() ; i++) {	30✔
184	error = (value(times_[i]) - blackVols_[i]);	27✔
185	squaredError += error * error * weights_[i];	27✔
186	}
187	return std::sqrt(n*squaredError/(n-1));	3✔
188	}
189
190	Real AbcdCalibration::maxError() const {	×
191	Real error, maxError = QL_MIN_REAL;	×
192	for (Size i=0; i<times_.size() ; i++) {	×
193	error = std::fabs(value(times_[i]) - blackVols_[i]);	×
194	maxError = std::max(maxError, error);	×
195	}
196	return maxError;	×
197	}
198
199	// calculate weighted differences
200	Array AbcdCalibration::errors() const {	17✔
201	Array results(times_.size());	17✔
202	for (Size i=0; i<times_.size() ; i++) {	170✔
203	results[i] = (value(times_[i]) - blackVols_[i])* std::sqrt(weights_[i]);	153✔
204	}
205	return results;	17✔
206	}
207
208	EndCriteria::Type AbcdCalibration::endCriteria() const{	1✔
209	return abcdEndCriteria_;	1✔
210	}
211
212	}

lballabio / QuantLib / 27603489947

Source File Press 'n' to go to next uncovered line, 'b' for previous

Source File
Press 'n' to go to next uncovered line, 'b' for previous