Problem_MT Class Reference

Constrained optimization problem. More...

#include <qle/math/problem_mt.hpp>

Public Member Functions
	Problem_MT (const std::vector< boost::shared_ptr< CostFunction >> &costFunctions, Constraint &constraint, const Array &initialValue=Array())
void	reset ()
Real	value (const Size i, const Array &x)
	call cost function i computation and increment evaluation counter
Array	values (const Size i, const Array &x)
	call cost values i computation and increment evaluation counter
void	gradient (const Size i, Array &grad_f, const Array &x)
	call cost function i gradient computation and increment
Real	valueAndGradient (const Size i, Array &grad_f, const Array &x)
	call cost function i computation and it gradient
Constraint &	constraint () const
	Constraint.
Integer	availableCostFunctions () const
	number of available independent cost functions
boost::shared_ptr< CostFunction >	costFunction (const Size i) const
	Cost function.
const std::vector< boost::shared_ptr< CostFunction > > &	costFunctions () const
	Cost funcionts.
void	setCurrentValue (const Array &currentValue)
const Array &	currentValue () const
	current value of the local minimum
void	setFunctionValue (Real functionValue)
Real	functionValue () const
	value of cost function
void	setGradientNormValue (Real squaredNorm)
Real	gradientNormValue () const
	value of cost function gradient norm
Integer	functionEvaluation () const
	number of evaluation of cost function
Integer	gradientEvaluation () const
	number of evaluation of cost function gradient

Protected Attributes
std::vector< boost::shared_ptr< CostFunction > >	costFunctions_
	Unconstrained cost function.
Constraint &	constraint_
	Constraint.
Array	currentValue_
	current value of the local minimum
Real	functionValue_
	function and gradient norm values at the curentValue_ (i.e. the last step)
Real	squaredNorm_
std::vector< Integer >	functionEvaluation_
	number of evaluation of cost function and its gradient
std::vector< Integer >	gradientEvaluation_

Detailed Description

Constrained optimization problem.

Warning:

The passed Constraint instances are stored by reference. The user of this class must make sure that they are not destroyed before the Problem instance.

Constructor & Destructor Documentation

Problem_MT()

Problem_MT	(	const std::vector< boost::shared_ptr< CostFunction >> &	costFunctions,
		Constraint &	constraint,
		const Array &	initialValue = Array()
	)

The requirement is that the cost functions can be evaluated from multiple threads without data races, which can e.g. be done by complete separation of the underlying data. Note that some methods in this class are not thread safe, namely reset, setCurrentValue, setFunctionValue, setGradientNormValue, i.e. those can not be used from several threads in a mt optimizer

Member Function Documentation

reset()

void reset

(

)

Warning:

it does not reset the current minumum to any initial value