golib  0.5
Public Types | Public Member Functions | Protected Attributes | List of all members
goMath::BarrierOptFunction< matrix_type, vector_type > Class Template Reference

An OptFunction that is used in BarrierOpt. More...

#include <gobarrieropt.h>

Public Types

typedef OptFunction< matrix_type, vector_type >::value_type value_type
 
typedef OptFunction< matrix_type, vector_type > function_type
 
- Public Types inherited from goMath::OptFunction< matrix_type, vector_type >
typedef vector_type::value_type value_type
 Value of the matrix and vector types matrix_type and vector_type.
 

Public Member Functions

 BarrierOptFunction (double eps=0.01)
 
void setT (value_type t)
 
value_type t () const
 
virtual value_type barrier (const vector_type &x)
 Barrier function $ \Phi $. More...
 
void setProblem (goAutoPtr< OptProblem< matrix_type, vector_type > > p)
 
virtual value_type operator() (const vector_type &x)
 Calculate $ t \cdot f(x) + \Phi(x) $. More...
 
virtual void grad (vector_type &x, vector_type &ret)
 Calculate the gradient of f at x. More...
 
virtual void hessian (vector_type &x, matrix_type &ret)
 
void barrierGradHessian (const vector_type &x, vector_type &grad, matrix_type &hess)
 
goAutoPtr< OptProblem< matrix_type, vector_type > > problem ()
 
- Public Member Functions inherited from goMath::OptFunction< matrix_type, vector_type >
virtual void hessian (vector_type &x, matrix_type &ret)
 Calculate the Hessian of f at x. More...
 

Protected Attributes

goAutoPtr< OptProblem< matrix_type, vector_type > > myProblem
 
vector_type myBufferGrad
 
matrix_type myBufferHess
 
value_type my_t
 

Additional Inherited Members

- Protected Member Functions inherited from goMath::OptFunction< matrix_type, vector_type >
 OptFunction (double eps=0.01)
 Constructor. More...
 

Detailed Description

template<class matrix_type, class vector_type>
class goMath::BarrierOptFunction< matrix_type, vector_type >

An OptFunction that is used in BarrierOpt.

This class provides gradient and Hessian methods for calculating the gradient and Hessian of an optimisation problem set with setProblem() by augmenting the gradient and Hessian of the original objective function by the ones induced by the log barrier function.

The gradient is calculated as ... FIXME

Member Function Documentation

◆ barrier()

template<class matrix_type , class vector_type >
virtual value_type goMath::BarrierOptFunction< matrix_type, vector_type >::barrier ( const vector_type &  x)
inlinevirtual

Barrier function $ \Phi $.

Calculates the barrier function

\[ \Phi(x) = -\sum_{i=1}^{m} \log (-f_i(x)) \]

for all inequality constraints $ f_i(x) \leq 0 $ which obviously must be met strictly.

Parameters
xPoint at which to calculate $ \Phi(x) $.

Reimplemented in goMath::BarrierOptFunctionPhase1< matrix_type, vector_type >.

◆ grad()

template<class matrix_type , class vector_type >
virtual void goMath::BarrierOptFunction< matrix_type, vector_type >::grad ( vector_type &  x,
vector_type &  ret 
)
inlinevirtual

Calculate the gradient of f at x.

Does a numerical approximation with forward differences. Re-implement for specialisations.

Note
x is modified during the call, but restored when the call returns.
Parameters
xPoint at which to calculate the gradient
retContains $\nabla f(x)$ on return

Reimplemented from goMath::OptFunction< matrix_type, vector_type >.

Reimplemented in goMath::BarrierOptFunctionPhase1< matrix_type, vector_type >.

◆ operator()()

template<class matrix_type , class vector_type >
virtual value_type goMath::BarrierOptFunction< matrix_type, vector_type >::operator() ( const vector_type &  x)
inlinevirtual

Calculate $ t \cdot f(x) + \Phi(x) $.

See also
barrier()

$ f(x) $ is the original function to be minimised, $ \Phi(x) $ is the log barrier.

Returns
$ t \cdot f(x) + \Phi(x) $

Implements goMath::OptFunction< matrix_type, vector_type >.

Reimplemented in goMath::BarrierOptFunctionPhase1< matrix_type, vector_type >.

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