LmiOracleBase< Arr036, Mat > Class Template Reference

Shared skeleton for Linear Matrix Inequality oracles. More...

#include <lmi_oracle_base.hpp>

Inheritance diagram for LmiOracleBase< Arr036, Mat >:

Protected Types
using	Cut = std::pair< Arr036, double >

Protected Member Functions
template<typename LDLT , typename Fn >
auto	assess_impl (LDLT &mgr, const std::vector< Mat > &F, const int sign, const Arr036 &x, Fn &&getA) -> Cut *
	Shared assess_feas skeleton: factor, witness, sym_quad, pack cut.

Protected Attributes
std::unique_ptr< Cut >	cut = std::make_unique<Cut>()
	Storage for the cut returned by assess_feas.

Detailed Description

template<typename Arr036, typename Mat = Arr036>
class LmiOracleBase< Arr036, Mat >

Shared skeleton for Linear Matrix Inequality oracles.

The three LMI oracle flavors (LmiOracle, Lmi0Oracle, LmiOldOracle) differ only in how the matrix elements A(i,j) are assembled (lazy lambda vs. eagerly-built matrix), in the sym_quad sign convention, and in the presence of a constant term B. The factorization / witness / cut-packing pipeline is identical and lives here.

Note

Template Method pattern: assess_impl is the fixed algorithm skeleton (factor -> witness -> sym_quad -> pack cut); each concrete oracle supplies its own getA callable, sign, and manager via the parameter list, keeping every public constructor and member intact.

Template Parameters

Arr036	Array type for the decision variables
Mat	Matrix type (defaults to Arr036)

Member Typedef Documentation

Cut

template<typename Arr036 , typename Mat = Arr036>

using LmiOracleBase< Arr036, Mat >::Cut = std::pair<Arr036, double>

protected

Member Function Documentation

assess_impl()

template<typename Arr036 , typename Mat = Arr036>

template<typename LDLT , typename Fn >

auto LmiOracleBase< Arr036, Mat >::assess_impl ( LDLT &  mgr, const std::vector< Mat > &  F, const int  sign, const Arr036 &  x, Fn &&  getA  ) -> Cut*

inlineprotected

Shared assess_feas skeleton: factor, witness, sym_quad, pack cut.

Template Parameters

LDLT	LDL^T manager type (LDLTMgr)
Fn	Callable with signature double(size_t, size_t)

Parameters

[in,out]	mgr	LDL^T factorization manager
[in]	F	Vector of matrices F_i (fixed problem data)
[in]	sign	+1 or -1 convention for sym_quad
[in]	x	Evaluation point
[in]	getA	Lazy accessor for matrix element A(i, j)

Returns

Cut* pointer to the packed cut, or nullptr if feasible

Member Data Documentation

cut

template<typename Arr036 , typename Mat = Arr036>

std::unique_ptr<Cut> LmiOracleBase< Arr036, Mat >::cut = std::make_unique<Cut>()

protected

Storage for the cut returned by assess_feas.

---

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

• ellalgo/oracles/lmi_oracle_base.hpp