Package org.hipparchus.optim.nonlinear.vector.constrained

Class ADMMQPKKT

java.lang.Object

org.hipparchus.optim.nonlinear.vector.constrained.ADMMQPKKT

All Implemented Interfaces:

KarushKuhnTuckerSolver<ADMMQPSolution>, OptimizationData

public class ADMMQPKKT
extends Object
implements KarushKuhnTuckerSolver<ADMMQPSolution>

Alternative Direction Method of Multipliers Solver.

Since:

3.1

Method Summary

Modifier and Type	Method	Description
void	initialize(RealMatrix newH, RealMatrix newA, RealVector newQ, int me, RealVector newLb, RealVector newUb, double rho, double newSigma, double newAlpha)	Initialize problem
ADMMQPSolution	iterate(RealVector... previousSol)	Iterate Karush–Kuhn–Tucker equation from given list of Vector
ADMMQPSolution	solve(RealVector b1, RealVector b2)	Solve Karush–Kuhn–Tucker equation from given right hand value.
void	updateSigmaRho(double newSigma, int me, double rho)	Update steps

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Method Detail

solve

public ADMMQPSolution solve(RealVector b1,
                            RealVector b2)

Solve Karush–Kuhn–Tucker equation from given right hand value.

Specified by:

solve in interface KarushKuhnTuckerSolver<ADMMQPSolution>

Parameters:

b1 - first right hand vector

b2 - second right hand vector

Returns:

Tuple with the solution x,Lambda,value

updateSigmaRho

public void updateSigmaRho(double newSigma,
                           int me,
                           double rho)

Update steps

Parameters:

newSigma - new regularization term sigma for Karush–Kuhn–Tucker solver

me - number of equality constraints

rho - new step size

initialize

public void initialize(RealMatrix newH,
                       RealMatrix newA,
                       RealVector newQ,
                       int me,
                       RealVector newLb,
                       RealVector newUb,
                       double rho,
                       double newSigma,
                       double newAlpha)

Initialize problem

Parameters:

newH - square matrix of weights for quadratic term

newA - constraints coefficients matrix

newQ - TBD

me - number of equality constraints

newLb - lower bound

newUb - upper bound

rho - step size

newSigma - regularization term sigma for Karush–Kuhn–Tucker solver

newAlpha - alpha filter for ADMM iteration

iterate

public ADMMQPSolution iterate(RealVector... previousSol)

Iterate Karush–Kuhn–Tucker equation from given list of Vector

Specified by:

iterate in interface KarushKuhnTuckerSolver<ADMMQPSolution>

Parameters:

previousSol - list of vectors

Returns:

Tuple with the solution x,Lambda,value