Package de.labathome.trustconstr

Class VectorFunction.VectorFunctionFactory

java.lang.Object

de.labathome.trustconstr.VectorFunction.VectorFunctionFactory

Enclosing class:

VectorFunction

public static class VectorFunction.VectorFunctionFactory
extends Object

Builder for VectorFunction.

Constructor Summary

Constructors

Constructor	Description
VectorFunctionFactory()	Public so callers can construct a fresh factory per build instead of mutating the shared VectorFunction.FACTORY singleton.

Method Summary

Modifier and Type	Method	Description
VectorFunction	build()
VectorFunction.VectorFunctionFactory	finiteDiffBounds(FiniteDifferenceBounds finiteDiffBounds)	Bounds within which FD perturbations must stay.
VectorFunction.VectorFunctionFactory	finiteDiffJacSparsity(Matrix finiteDiffJacSparsity)	Specify a Jacobian sparsity pattern: a matrix with non-zero elements exactly where non-zero Jacobian entries are expected.
VectorFunction.VectorFunctionFactory	finiteDiffRelStep(Matrix finiteDiffRelStep)	Per-component relative step size for the FD Jacobian/Hessian.
VectorFunction.VectorFunctionFactory	fun(Function<Matrix,Matrix> fun)	Set the vector-valued function to evaluate.
VectorFunction.VectorFunctionFactory	hess(FiniteDifferenceMethod hess)	Compute the Hessian by finite differences instead of analytically.
VectorFunction.VectorFunctionFactory	hess(HessianUpdateStrategy hess)	Approximate the Hessian via a quasi-Newton update strategy (e.g.
VectorFunction.VectorFunctionFactory	hess(BiFunction<Matrix,Matrix,Matrix> hess)	Provide an analytic constraint Hessian-of-Lagrangian (x, v) -> Sum v[i] H_{c_i}(x).
VectorFunction.VectorFunctionFactory	jac(FiniteDifferenceMethod jac)	Compute the Jacobian by finite differences instead of analytically.
VectorFunction.VectorFunctionFactory	jac(Function<Matrix,Matrix> jac)	Provide an analytic Jacobian.
VectorFunction.VectorFunctionFactory	sparseJacobian()	Force the Jacobian return type to a sparse representation.
VectorFunction.VectorFunctionFactory	x0(Matrix x0)	Provide the initial point.

Methods inherited from class java.lang.Object

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

Constructor Details

VectorFunctionFactory

public VectorFunctionFactory()

Public so callers can construct a fresh factory per build instead of mutating the shared VectorFunction.FACTORY singleton. Same fix pattern as ScalarFunction.ScalarFunctionFactory and the BFGS/SR1 factories.

Method Details

fun

public VectorFunction.VectorFunctionFactory fun(Function<Matrix,Matrix> fun)

Set the vector-valued function to evaluate.

Parameters:

fun - function R^n -> R^m

Returns:

this factory, for chaining

x0

public VectorFunction.VectorFunctionFactory x0(Matrix x0)

Provide the initial point.

Parameters:

x0 - starting iterate (n x 1)

Returns:

this factory, for chaining

jac

public VectorFunction.VectorFunctionFactory jac(Function<Matrix,Matrix> jac)

Provide an analytic Jacobian.

Parameters:

jac - analytic Jacobian R^n -> R^{m x n}

Returns:

this factory, for chaining

Throws:

RuntimeException - if a Jacobian was already specified

jac

public VectorFunction.VectorFunctionFactory jac(FiniteDifferenceMethod jac)

Compute the Jacobian by finite differences instead of analytically.

Parameters:

jac - FD scheme to use

Returns:

this factory, for chaining

Throws:

RuntimeException - if a Jacobian was already specified

finiteDiffJacSparsity

public VectorFunction.VectorFunctionFactory finiteDiffJacSparsity(Matrix finiteDiffJacSparsity)

Specify a Jacobian sparsity pattern: a matrix with non-zero elements exactly where non-zero Jacobian entries are expected. Enables Curtis-Powell-Reid column grouping for the FD path.

Parameters:

finiteDiffJacSparsity - sparsity pattern (m x n)

Returns:

this factory, for chaining

sparseJacobian

public VectorFunction.VectorFunctionFactory sparseJacobian()

Force the Jacobian return type to a sparse representation.

Returns:

this factory, for chaining

hess

public VectorFunction.VectorFunctionFactory hess(BiFunction<Matrix,Matrix,Matrix> hess)

Provide an analytic constraint Hessian-of-Lagrangian (x, v) -> Sum v[i] H_{c_i}(x).

Parameters:

hess - analytic Hessian-of-Lagrangian

Returns:

this factory, for chaining

Throws:

RuntimeException - if a Hessian was already specified

hess

public VectorFunction.VectorFunctionFactory hess(FiniteDifferenceMethod hess)

Compute the Hessian by finite differences instead of analytically. Cannot be combined with an FD Jacobian.

Parameters:

hess - FD scheme to use for the Hessian

Returns:

this factory, for chaining

Throws:

RuntimeException - if a Hessian was already specified

hess

public VectorFunction.VectorFunctionFactory hess(HessianUpdateStrategy hess)

Approximate the Hessian via a quasi-Newton update strategy (e.g. BFGS, SR1).

Parameters:

hess - Hessian-update strategy instance

Returns:

this factory, for chaining

Throws:

RuntimeException - if a Hessian was already specified

finiteDiffRelStep

public VectorFunction.VectorFunctionFactory finiteDiffRelStep(Matrix finiteDiffRelStep)

Per-component relative step size for the FD Jacobian/Hessian. May be null to derive a default from machine epsilon.

Parameters:

finiteDiffRelStep - relative step size (n x 1); may be null

Returns:

this factory, for chaining

finiteDiffBounds

public VectorFunction.VectorFunctionFactory finiteDiffBounds(FiniteDifferenceBounds finiteDiffBounds)

Bounds within which FD perturbations must stay. Defaults to unbounded.

Parameters:

finiteDiffBounds - box bounds for FD perturbations

Returns:

this factory, for chaining

build

public VectorFunction build()

Returns:

a fully-configured VectorFunction

Throws:

RuntimeException - if neither analytic nor FD Jacobian/Hessian was set