Algorithm Class Reference

This is the main algorithm base class; actual algorithms should be derived. More...

#include <Algorithm.hpp>

Inheritance diagram for Algorithm:

List of all members.

Public Member Functions
	Algorithm ()
	Default constructor; doesn't initialize much, so setup() must be called.
virtual void	setup (const ptree &all, const string &outputdirname)
	Setup case-specific settings.
virtual void	run ()
	Run the algorithm as the setup tells us to.
Protected Member Functions
virtual void	impl_setup (const ptree &all)=0
	Derived classes must know how to setup their additional data structures.
virtual size_t	nextID ()
	Give the next available ID (running index).
void	markAsParents (size_t childID, const vector< size_t > &parentIDs)
	Remember a parenting relationship, for tracing the runs.
void	traceMeasures (size_t childID, const vector< double > &measures)
	Trace some measures of an individual, possibly more than just objectives.
void	writeParentHistory (ostream &os) const
	Write the traced parenting history to a stream.
void	writeTracedMeasures (ostream &os) const
	Write the traced measures to a stream.
Protected Attributes
size_t	_maxiter
	Number of generations to carry out.
string	_outputdirname
	Output directory in which all the results and traces will be saved.
size_t	_nextID = 1
	Next ID to give to a modified individual (unique running index).
vector< vector< size_t > >	_parentsOf
	The IDs of parents; for tracing the algorithm progression.
vector< vector< double > >	_tracedMeasures
	Traced measures; for tracing the algorithm progression.

---

Detailed Description

This is the main algorithm base class; actual algorithms should be derived.

This base class offers the general framework for setup and iterative running, but knows nothing about individual representation or operators actually used. This also handles the "dirty and mundane work" of logging outputs etc. Formatting can, of course, be done only by derived classes that know about the actual representation.

Boost property_tree is used for setup, because of its nice parser for structured configuration.

---

Constructor & Destructor Documentation

Algorithm::Algorithm

(

)

Default constructor; doesn't initialize much, so setup() must be called.

---

Member Function Documentation

void Algorithm::markAsParents	(	size_t	childID,
		const vector< size_t > &	parentIDs
	)			[protected]

Remember a parenting relationship, for tracing the runs.

void Algorithm::run

(

)

[virtual]

Run the algorithm as the setup tells us to.

Reimplemented in MlpLearnAlgorithm.

void Algorithm::setup	(	const ptree &	all,
		const string &	outputdirname
	)			[virtual]

Setup case-specific settings.

void Algorithm::traceMeasures	(	size_t	childID,
		const vector< double > &	measures
	)			[protected]

Trace some measures of an individual, possibly more than just objectives.

void Algorithm::writeParentHistory

(

ostream &

os

)

const [protected]

Write the traced parenting history to a stream.

void Algorithm::writeTracedMeasures

(

ostream &

os

)

const [protected]

Write the traced measures to a stream.

---

Member Data Documentation

size_t Algorithm::_maxiter [protected]

Number of generations to carry out.

size_t Algorithm::_nextID = 1 [protected]

Next ID to give to a modified individual (unique running index).

string Algorithm::_outputdirname [protected]

Output directory in which all the results and traces will be saved.

vector< vector <size_t> > Algorithm::_parentsOf [protected]

The IDs of parents; for tracing the algorithm progression.

vector< vector <double> > Algorithm::_tracedMeasures [protected]

Traced measures; for tracing the algorithm progression.

---

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

• src/Algorithm.hpp
• src/Algorithm.cpp