Base class for steps that update penalty parameters based on the Lagrange multipliers lambda_k (or some approximation to them). More...

Inheritance diagram for MoochoPack::MeritFunc_PenaltyParamUpdate_AddedStep:

Pure virtual methods to be overridden by subclasses
virtual void	small_mu (value_type small_mu)=0
	Set the smallest value a penalty parameter is allowed to be. More...

virtual value_type	small_mu () const =0

virtual void	min_mu_ratio (value_type min_mu_ratio)
	Set the ratio of min(mu(i))/max(mu(i)) >= min_mu_ratio. More...

virtual value_type	min_mu_ratio () const

virtual void	mult_factor (value_type mult_factor)=0
	Set set the factor for mu = (1 + mult_factor) * abs(lambda_k(i)). More...

virtual value_type	mult_factor () const =0

virtual void	kkt_near_sol (value_type kkt_near_sol)=0
	Set the total KKT error ( max(\|\|rGL\|\|inf/max(1,\|\|Gf\|\|inf),\|\|c\|\|inf) ) above which the penalty parameter will be allowed to be reduced. More...

virtual value_type	kkt_near_sol () const =0

Detailed Description

Base class for steps that update penalty parameters based on the Lagrange multipliers lambda_k (or some approximation to them).

This class contains methods for setting and querying values that determine how the penalty parameters are updated.

Member Function Documentation

virtual void MoochoPack::MeritFunc_PenaltyParamUpdate_AddedStep::small_mu ( value_type small_mu )

pure virtual

Set the smallest value a penalty parameter is allowed to be.

virtual value_type MoochoPack::MeritFunc_PenaltyParamUpdate_AddedStep::small_mu ( ) const

pure virtual

virtual void MoochoPack::MeritFunc_PenaltyParamUpdate_AddedStep::min_mu_ratio ( value_type min_mu_ratio )

inlinevirtual

Set the ratio of min(mu(i))/max(mu(i)) >= min_mu_ratio.

If there is only one penalty parameter this is ignored. The default implementation is to just return 1.

virtual value_type MoochoPack::MeritFunc_PenaltyParamUpdate_AddedStep::min_mu_ratio ( ) const

inlinevirtual

virtual void MoochoPack::MeritFunc_PenaltyParamUpdate_AddedStep::mult_factor ( value_type mult_factor )

pure virtual

Set set the factor for mu = (1 + mult_factor) * abs(lambda_k(i)).

Here it is expented that mult_factor will be very small (i.e. 1e-4)

virtual value_type MoochoPack::MeritFunc_PenaltyParamUpdate_AddedStep::mult_factor ( ) const

pure virtual

virtual void MoochoPack::MeritFunc_PenaltyParamUpdate_AddedStep::kkt_near_sol ( value_type kkt_near_sol )

pure virtual

Set the total KKT error ( max(||rGL||inf/max(1,||Gf||inf),||c||inf) ) above which the penalty parameter will be allowed to be reduced.

If the KKT error is less than near_solution the penalty parameter will only be increased. This is usually needed to prove convergence.

virtual value_type MoochoPack::MeritFunc_PenaltyParamUpdate_AddedStep::kkt_near_sol ( ) const

pure virtual

The documentation for this class was generated from the following file: