Decomposition Class Reference

Decompose a function or problem. More...

#include <decomp.h>

Collaboration diagram for Decomposition:

Collaboration graph
[legend]
List of all members.

Public Member Functions

 Decomposition (Pointer< Param > param_=NULL)
 (Default-)Constructor.
void set_sparsity_pattern (MinlpProblem &prob, const vector< vector< dvector > > &sample_set)
Pointer< MinlpProblemdecompose (MinlpProblem &prob, vector< vector< dvector > > &sample_set)
 Decompose a problem.
Pointer< MinlpProblemdecompose (MinlpProblem &prob, vector< Pointer< SplittingScheme2 > > &ss)
void decompose (SepQcFunc &f, int block_offset, const SepQcFunc &old_f, int k, const SplittingScheme2 &ss, Pointer< dvector > primal)
Pointer< SplittingScheme2get_splittingscheme (MinlpProblem &prob, int blocknr)
Pointer< SepQcFuncdecompose (Pointer< Func > old_f, const dvector &primal)

Public Attributes

Pointer< set< int > > I_var

Private Attributes

bool replace_if_quadratic
 Indicates, whether a (s-)function should be replaced, if it seems to be quadratic (constant hessian).
int avg_blocksize
 The average block size, we should try to achieve.
Pointer< Paramparam
 Parameters.

Detailed Description

Decompose a function or problem.

Constructs also sparsity pattern and computes nonlinear and nonquadratic variable indices.

Parameters:
Decomposition replace quadratic options 0 or 1 default 1 If 1, (s-)functions with constant hessians are replaced by their quadratic approximation.
Decomposition average blocksize options integer $ 1$ default 5 level 1 The average blocksize, we should try to partition the sparsity graph into.

Definition at line 261 of file decomp.h.


Constructor & Destructor Documentation

Decomposition::Decomposition ( Pointer< Param param_ = NULL  )  [inline]

(Default-)Constructor.

Parameters:
replace_if_quadratic_ Should we replace a s-function, if it's Hessian is constant over some sample point? Default is false.

Definition at line 281 of file decomp.h.


Member Function Documentation

void Decomposition::set_sparsity_pattern ( MinlpProblem prob,
const vector< vector< dvector > > &  sample_set 
)

Pointer<MinlpProblem> Decomposition::decompose ( MinlpProblem prob,
vector< vector< dvector > > &  sample_set 
)

Decompose a problem.

Computes sparsity patterns and indices of quadratic and nonquadratic variables. Decompose sparsity-graphs. Constructs SplittingScheme. Decompose problem according to this information.

Parameters:
ss A reference to a Pointer, where we can store the SplittingScheme in.
prob The problem to decompose.
sample_set Points to compute the hessian for to get sparsity pattern, needed for functions, where s[k]!=NULL.
Returns:
The decomposed problem.

Pointer<MinlpProblem> Decomposition::decompose ( MinlpProblem prob,
vector< Pointer< SplittingScheme2 > > &  ss 
)

void Decomposition::decompose ( SepQcFunc f,
int  block_offset,
const SepQcFunc old_f,
int  k,
const SplittingScheme2 ss,
Pointer< dvector primal 
)

Pointer<SplittingScheme2> Decomposition::get_splittingscheme ( MinlpProblem prob,
int  blocknr 
)

Pointer<SepQcFunc> Decomposition::decompose ( Pointer< Func old_f,
const dvector primal 
)


Member Data Documentation

bool Decomposition::replace_if_quadratic [private]

Indicates, whether a (s-)function should be replaced, if it seems to be quadratic (constant hessian).

Definition at line 265 of file decomp.h.

int Decomposition::avg_blocksize [private]

The average block size, we should try to achieve.

Definition at line 269 of file decomp.h.

Pointer<Param> Decomposition::param [private]

Parameters.

Definition at line 273 of file decomp.h.

Pointer<set<int> > Decomposition::I_var

Definition at line 276 of file decomp.h.


The documentation for this class was generated from the following file:
Generated on Mon Oct 20 03:12:26 2008 for LaGO by  doxygen 1.4.7