Meta::MOptRun Class Reference

Uses genetic algorithms to perform an optimization run. More...

#include <MOptRun.h>

List of all members.

Public Member Functions
	MOptRun ()
virtual	~MOptRun ()
virtual void	run ()
void	minimize (const mvec &start, const mtime &timeout, size_t population=100, size_t samples=1, double target=mnum::min())
void	maximize (const mvec &start, const mtime &timeout, size_t population=100, size_t samples=1, double target=mnum::max())
Static Public Member Functions
static void	finish (double metric)

---

Detailed Description

Uses genetic algorithms to perform an optimization run.

A flexible interface for performing an optimization run using a genetic algorithms approach. This class is extended to provide a run() method that performs an application-specific operation involving its input iteration parameters and judges the performance by computing a metric that is then passed to finish(). The goal of the optimization is to either maximize or minimize this metric. The optimization process runs several trials until either a target metric is achieved or maxTime has expired. Arguments are passed to each run through the standard MProgram arguments interface.

---

Constructor & Destructor Documentation

Meta::MOptRun::MOptRun

(

)

Constructor.

virtual Meta::MOptRun::~MOptRun

(

)

[virtual]

Destructor.

---

Member Function Documentation

static void Meta::MOptRun::finish

(

double

metric

)

[static]

There should exist a stopping point in run() whereby a metric is computed and finish() is called.

void Meta::MOptRun::maximize	(	const mvec &	start,
		const mtime &	timeout,
		size_t	population = 100,
		size_t	samples = 1,
		double	target = mnum::max()
	)

Similar to minimize(), start the optimization run where the goal is to maximize the metric.

void Meta::MOptRun::minimize	(	const mvec &	start,
		const mtime &	timeout,
		size_t	population = 100,
		size_t	samples = 1,
		double	target = mnum::min()
	)

Called once to start the optimization run where the goal is to minimize the metric.

Parameters:

	start	A vector containing a sequence of named mappings to the start parameters, e.g: [[x:2,y:[1,2.4,3]],[x:3,y:[5,6.3,7,3]]]. All values must be derived of numerical types and each must define a consistent structure.
	timeout	Stop the run if this timeout is reached.
	population	The parameter set population that the run uses -- not necessarily the size of the start vector.
	samples	Useful for stochastic runs, where multiple iterations are performed for the same input parameters with the metrics that finish() accepts averaged.
	target	Halt when this metric is achieved.

virtual void Meta::MOptRun::run

(

)

[virtual]

Application-specific method that must be re-implemented to perform a run iteration.

---

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

• AndroMeta/include/Meta/MOptRun.h