LinearizedConCutGenerator Class Reference

#include <cuts.h>

Collaboration diagram for LinearizedConCutGenerator:

Public Member Functions
	LinearizedConCutGenerator (Pointer< MinlpProblem > prob_, Pointer< MINLPData > minlpdata_=NULL, Pointer< Reformulation > reform_=NULL)
LinearizationCut	get_cut (const dvector &x, int c, int block_nr, double val=INFINITY)
	Returns a linearization of a constraint, which is defined over one block only.
int	get_cuts (list< pair< LinearizationCut, pair< int, bool > > > &cuts, const MINLPData::ObjCon &objcon, int objcon_nr, const dvector &x, const dvector &lower, const dvector &upper, bool violated_polyest_only=false)
int	get_cuts (list< pair< LinearizationCut, pair< int, bool > > > &cuts, const MINLPData::Constraint &objcon, int objcon_nr, const dvector &x, const dvector &lower, const dvector &upper, bool violated_polyest_only=false)
int	get_cuts (list< pair< LinearizationCut, pair< int, bool > > > &cuts, const dvector &x, const dvector &lower, const dvector &upper, bool violated_polyest_only=false)
	Computes linearization cuts by linearization of the convex relaxation in a given point.
Pointer< LinearizationCut >	update_cut (LinearizationCut &cut, unsigned int block_nr, const dvector &lower, const dvector &upper)
	Updates a cut, using a current box.
void	set_problem (Pointer< MinlpProblem > prob_)
void	set_reform (Pointer< Reformulation > reform_)
void	set_MINLPData (Pointer< MINLPData > minlpdata_)
Public Attributes
double	max_violation
	The maximum violation of a quadratic underestimator in a reference point.
Private Attributes
Pointer< MinlpProblem >	prob
Pointer< MINLPData >	minlpdata
Pointer< Reformulation >	reform
Static Private Attributes
static const double	tol

Detailed Description

Definition at line 389 of file cuts.h.

Constructor & Destructor Documentation

LinearizedConCutGenerator::LinearizedConCutGenerator	(	Pointer< MinlpProblem >	prob_,
		Pointer< MINLPData >	minlpdata_ = `NULL`,
		Pointer< Reformulation >	reform_ = `NULL`
	)

Member Function Documentation

LinearizationCut LinearizedConCutGenerator::get_cut	(	const dvector &	x,
		int	c,
		int	block_nr,
		double	val = `INFINITY`
	)

Returns a linearization of a constraint, which is defined over one block only.

Parameters:

	x	The point, in which we linearize the constraint. block[block_nr].size()
	c	The constraint number.
	block_nr	The number of the block, for which the constraint is defined.
	val	The value of the constraint in x, if available. INFINITY else.

int LinearizedConCutGenerator::get_cuts	(	list< pair< LinearizationCut, pair< int, bool > > > &	cuts,
		const MINLPData::ObjCon &	objcon,
		int	objcon_nr,
		const dvector &	x,
		const dvector &	lower,
		const dvector &	upper,
		bool	violated_polyest_only = `false`
	)

int LinearizedConCutGenerator::get_cuts	(	list< pair< LinearizationCut, pair< int, bool > > > &	cuts,
		const MINLPData::Constraint &	objcon,
		int	objcon_nr,
		const dvector &	x,
		const dvector &	lower,
		const dvector &	upper,
		bool	violated_polyest_only = `false`
	)

int LinearizedConCutGenerator::get_cuts	(	list< pair< LinearizationCut, pair< int, bool > > > &	cuts,
		const dvector &	x,
		const dvector &	lower,
		const dvector &	upper,
		bool	violated_polyest_only = `false`
	)

Computes linearization cuts by linearization of the convex relaxation in a given point.

Parameters:

	cuts	A list of type { cut , { blocknr, is the function where the cut was derived from convex? } } to store the cuts.
	x	The point where to compute the cut.
	lower	The current lower bound on the variables.
	upper	The current upper bound on the variables.
	violated_polyest_only	If true, cuts are only generated for functions where x violates the polynomial under/over estimator. Useful if cut is ment to cuf off a solution of the relaxation.

Pointer<LinearizationCut> LinearizedConCutGenerator::update_cut	(	LinearizationCut &	cut,
		unsigned int	block_nr,
		const dvector &	lower,
		const dvector &	upper
	)

Updates a cut, using a current box.

Parameters:

	lower	The lower bounds on the variables, here only the lower bounds for the considered block.
	upper	The upper boudns on the variables, here only the upper bounds for the considered block.

void LinearizedConCutGenerator::set_problem ( Pointer< MinlpProblem > prob_ ) [inline]