slp::ProblemScaling< Scalar > Struct Template Reference

#include </home/runner/work/Sleipnir/Sleipnir/include/sleipnir/optimization/solver/util/problem_scaling.hpp>

Public Types
using	DenseVector = Eigen::Vector< Scalar, Eigen::Dynamic >
	Type alias for dense vector.
using	SparseMatrix = Eigen::SparseMatrix< Scalar >
	Type alias for sparse matrix.
using	SparseVector = Eigen::SparseVector< Scalar >
	Type alias for sparse vector.

Public Member Functions
	ProblemScaling ()=default
	Constructs identity problem scaling.
	ProblemScaling (Scalar f, const DenseVector &c_e, const DenseVector &c_i)
	ProblemScaling (const DenseVector &g)
	ProblemScaling (const DenseVector &g, const SparseMatrix &A_e)
	ProblemScaling (const DenseVector &g, const SparseMatrix &A_e, const SparseMatrix &A_i)
bool	is_identity () const

Public Attributes
Scalar	f = Scalar(1)
	Cost scaling factor d_f.
DenseVector	c_e
	Equality constraint scaling factors d_cₑ.
DenseVector	c_i
	Inequality constraint scaling factors d_cᵢ.

Detailed Description

template<typename Scalar>
struct slp::ProblemScaling< Scalar >

Automatic problem scaling factors.

Template Parameters

Scalar

Scalar type.

Constructor & Destructor Documentation

ProblemScaling() [1/4]

template<typename Scalar >

slp::ProblemScaling< Scalar >::ProblemScaling ( Scalar  f, const DenseVector &  c_e, const DenseVector &  c_i  )

inline

Constructs problem scaling from explicit factors.

Parameters

f	Cost scaling factor d_f.
c_e	Equality constraint scaling factors d_cₑ.
c_i	Inequality constraint scaling factors d_cᵢ.

ProblemScaling() [2/4]

template<typename Scalar >

slp::ProblemScaling< Scalar >::ProblemScaling ( const DenseVector &  g )

inlineexplicit

Computes Newton problem scaling.

Scales the cost so the largest gradient component at the starting point is at most gₘₐₓ:

d_f = min(1, gₘₐₓ / ‖∇f(x₀)‖_∞)

See §3.8 Automatic Scaling of the Problem Statement in [2].

Parameters

g	Cost gradient ∇f, evaluated at the starting point.

ProblemScaling() [3/4]

template<typename Scalar >

slp::ProblemScaling< Scalar >::ProblemScaling ( const DenseVector &  g, const SparseMatrix &  A_e  )

inline

Computes SQP problem scaling.

Scales the cost and each equality constraint so the largest gradient component at the starting point is at most gₘₐₓ:

d_f = min(1, gₘₐₓ / ‖∇f(x₀)‖_∞) d_cₑ[j] = min(1, gₘₐₓ / ‖∇cₑⱼ(x₀)‖_∞)

See §3.8 Automatic Scaling of the Problem Statement in [2].

Parameters

g	Cost gradient ∇f, evaluated at the starting point.
A_e	Equality constraint Jacobian Aₑ, evaluated at the starting point.

ProblemScaling() [4/4]

template<typename Scalar >

slp::ProblemScaling< Scalar >::ProblemScaling ( const DenseVector &  g, const SparseMatrix &  A_e, const SparseMatrix &  A_i  )

inline

Computes interior-point problem scaling.

Scales the cost and each constraint so the largest gradient component at the starting point is at most gₘₐₓ:

d_f = min(1, gₘₐₓ / ‖∇f(x₀)‖_∞) d_c[j] = min(1, gₘₐₓ / ‖∇cⱼ(x₀)‖_∞)

See §3.8 Automatic Scaling of the Problem Statement in [2].

Parameters

g	Cost gradient ∇f, evaluated at the starting point.
A_e	Equality constraint Jacobian Aₑ, evaluated at the starting point.
A_i	Inequality constraint Jacobian Aᵢ, evaluated at the starting point.

Member Function Documentation

is_identity()

template<typename Scalar >

bool slp::ProblemScaling< Scalar >::is_identity ( ) const

inline

Whether the problem scaling is identity.

Returns

True if the problem scaling is identity.

---

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

• /home/runner/work/Sleipnir/Sleipnir/include/sleipnir/optimization/solver/util/problem_scaling.hpp