ConjugateGradient (3) Linux Manual Page
QuantLib::ConjugateGradient – Multi-dimensional Conjugate Gradient class.
Synopsis
#include <ql/math/optimization/conjugategradient.hpp>Inherits QuantLib::LineSearchBasedMethod.
Public Member Functions
ConjugateGradient (const boost::shared_ptr< LineSearch > &lineSearch=boost::shared_ptr< LineSearch >()) virtual EndCriteria::Type minimize (Problem &P, const EndCriteria &endCriteria)
solve the optimization problem P
Detailed Description
Multi-dimensional Conjugate Gradient class.Fletcher-Reeves-Polak-Ribiere algorithm adapted from Numerical Recipes in C, 2nd edition.
User has to provide line-search method and optimization end criteria. Search direction $ d_i = – f'(x_i) + c_i*d_{i-1} $ where $ c_i = ||f'(x_i)||^2/||f'(x_{i-1})||^2 $ and $ d_1 = – f'(x_1) $