Package de.labathome.trustconstr

Class LagrangianHessian

java.lang.Object

de.labathome.trustconstr.LagrangianHessian

public final class LagrangianHessian
extends Object

Lagrangian Hessian H_obj(x) + Sum_i v_i * H_{c_i}(x) as a matrix-free LinearOperator-builder. Counterpart to scipy's LagrangianHessian class (_trustregion_constr/minimize_trustregion_constr.py:35).

Construct with the objective Hessian callable and a constraint Hessian-of-Lagrangian callable; invoke apply(Matrix, double[], double[]) to obtain a LinearOperator that evaluates (H_obj + H_constr) * p for a given p.

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static interface	LagrangianHessian.ConstraintHessian	Constraint Hessian-of-Lagrangian: Sum_i v_eq[i] * H_{eq,i}(x) + Sum_j v_ineq[j] * H_{ineq,j}(x).

Constructor Summary

Constructors

Constructor	Description
LagrangianHessian(int n, Function<Matrix,Matrix> objectiveHess, LagrangianHessian.ConstraintHessian constraintsHess)	Build a Lagrangian Hessian for an n-variable problem.

Method Summary

Modifier and Type	Method	Description
LinearOperator	apply(Matrix x, double[] vEq, double[] vIneq)
int	n()

Methods inherited from class java.lang.Object

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

Constructor Details

LagrangianHessian

public LagrangianHessian(int n,
 Function<Matrix,Matrix> objectiveHess,
 LagrangianHessian.ConstraintHessian constraintsHess)

Build a Lagrangian Hessian for an n-variable problem.

Parameters:

n - number of decision variables

objectiveHess - objective Hessian callable x -> H_obj(x)

constraintsHess - constraint Hessian-of-Lagrangian callable (x, vEq, vIneq) -> Sum_i v_i H_{c_i}(x); may return null when no constraint contributes (the apply() result is then just the objective Hessian)

Method Details

apply

public LinearOperator apply(Matrix x,
 double[] vEq,
 double[] vIneq)

Parameters:

x - current iterate (n x 1)

vEq - equality multipliers (may be null or empty)

vIneq - inequality multipliers (may be null or empty)

Returns:

matrix-free linear operator that applies (H_obj(x) + H_constr(x, vEq, vIneq)) * p to a given input vector p

n

public int n()

Returns:

number of decision variables this Hessian is sized for