IMP logo

IMP::gsl::Simplex Class Reference


Detailed Description

A simplex optimizer taken from GSL.

Simplex works by modifying a simplex in the space of the optimized attributes. The algorithm may not behave well when using score states, such as those involved in rigid bodies, which significantly change the values of the attributes. Then, again, it may work just fine. But be aware that it is not understood.

The main advantage of Simplex is that it is a local optimizer that does not require derivatives.

Inheritance diagram for IMP::gsl::Simplex:

Inheritance graph
[legend]

Public Member Functions

virtual std::string get_type_name () const
virtual ::IMP::VersionInfo get_version_info () const
virtual Float optimize (unsigned int max_steps)
 Simplex (Model *m=NULL)
Parameters
The parameters are:

  • minumum_size: the optimization stops when the size of the simplex falls below this. The size is defined as the average distance from the centroid to the simplex vertices.

  • initial_length: the length of the initial sizes of the simplex. Make sure that this covers the optimal solution, given the starting configuration.

Note:
Both quantities are relative to rescaled attribues and so should be numbers between 0 and 1.


void set_initial_length (double length)
void set_minimum_size (double d)

Friends

template<class T >
void IMP::internal::unref (T *)

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

Generated on Mon Mar 8 23:09:02 2010 for IMP by doxygen 1.5.8