Base object for Optimization methods. More...

Inheritance diagram for ObjCryst::OptimizationObj:

Classes
struct	LogLikelihoodStats
	Statistics about each object contributing to the overall Log(likelihood) More...

Public Member Functions
	OptimizationObj (const string name="")
	Constructor.
virtual	~OptimizationObj ()
	Destructor.
virtual void	RandomizeStartingConfig ()
	Randomize starting configuration.
virtual void	Optimize (long &nbSteps, const bool silent=false, const REAL finalcost=0, const REAL maxTime=-1)=0
	Launch optimization (a single run) for N steps.
virtual void	MultiRunOptimize (long &nbCycle, long &nbSteps, const bool silent=false, const REAL finalcost=0, const REAL maxTime=-1)=0
	Launch optimization for multiple runs of N steps.
void	FixAllPar ()
	Fix all parameters.
void	SetParIsFixed (const string &parName, const bool fix)
	Fix one parameter.
void	SetParIsFixed (const RefParType *type, const bool fix)
	Fix one family of parameters.
void	UnFixAllPar ()
	UnFix All parameters.
void	SetParIsUsed (const string &parName, const bool use)
	Set a parameter to be used.
void	SetParIsUsed (const RefParType *type, const bool use)
	Set a family of parameters to be used.
void	SetLimitsRelative (const string &parName, const REAL min, const REAL max)
	Change the relative limits for a parameter from its name.
void	SetLimitsRelative (const RefParType *type, const REAL min, const REAL max)
	Change the relative limits for a family of parameter.
void	SetLimitsAbsolute (const string &parName, const REAL min, const REAL max)
	Change the absolute limits for a parameter from its name.
void	SetLimitsAbsolute (const RefParType *type, const REAL min, const REAL max)
	Change the absolute limits for a family of parameter.
virtual REAL	GetLogLikelihood () const
	The optimized (minimized, actually) function.
void	StopAfterCycle ()
	Stop after the current cycle. USed for interactive refinement.
virtual void	DisplayReport ()
	Show report to the user during refinement. Used for GUI update.
void	AddRefinableObj (RefinableObj &)
	Add a refined object. All sub-objects are also added.
RefinableObj &	GetFullRefinableObj (const bool rebuild=true)
	Get the RefinableObj with all the parameters from all refined objects.
virtual void	XMLOutput (ostream &os, int indent=0) const =0
	Output a description of the object in XML format to a stream.
virtual void	XMLInput (istream &is, const XMLCrystTag &tag)=0
	Input in XML format from a stream, restoring the set of refined objects and the associated cost functions.
const string &	GetName () const
	Get the name for this object.
void	SetName (const string &)
	Set the name for this object.
virtual const string	GetClassName () const
	Get the name for this class type.
virtual void	Print () const
	Print some information about this object.
void	RestoreBestConfiguration ()
	Restore the Best configuration.
bool	IsOptimizing () const
	Are we busy optimizing ?
void	TagNewBestConfig ()
	During a global optimization, tell all objects that the current config is the latest "best" config.
REAL	GetLastOptimElapsedTime () const
	Get the elapsed time (in seconds) during the last optimization.
MainTracker &	GetMainTracker ()
	Get the MainTracker.
const MainTracker &	GetMainTracker () const
	Get the MainTracker.
RefObjOpt &	GetXMLAutoSaveOption ()
const RefObjOpt &	GetXMLAutoSaveOption () const
const REAL &	GetBestCost () const
	Access to current best cost.
REAL &	GetBestCost ()
	Access to current best cost.
virtual void	BeginOptimization (const bool allowApproximations=false, const bool enableRestraints=false)
	Begin optimization for all objects.
virtual void	EndOptimization ()
	End optimization for all objects.

Protected Member Functions
void	PrepareRefParList ()
virtual void	InitOptions ()
	Initialization of options.
virtual void	UpdateDisplay ()
	Update Display (if any display is available), when a new 'relevant' configuration is reached.
void	BuildRecursiveRefObjList ()
	(Re)build OptimizationObj::mRecursiveRefinedObjList, if an object has been added or modified.

Protected Attributes
RefinableObj	mRefParList
	The refinable par list used during refinement.
string	mName
	Name of the GlobalOptimization object.
string	mSaveFileName
	File name where refinement info is saved (NOT USED so far...)
long	mNbTrial
	Number of trials so far.
REAL	mBestCost
	Best value of the cost function so far.
long	mBestParSavedSetIndex
	Index of the 'best' saved parameter set.
unsigned long	mContext
	The current 'context', in the case the optimization is run in different parallel contexts.
map< unsigned long, map< const RefinableObj *, LogLikelihoodStats > >	mvContextObjStats
	Statistics for each context (mutable for dynamic update during optimization)
map< const RefinableObj *, DynamicObjWeight >	mvObjWeight
	Weights for each objects in each context (mutable for dynamic update during optimization)
std::vector< pair< long, REAL > >	mvSavedParamSet
	List of saved parameter sets.
bool	mIsOptimizing
	True if a refinement is being done. For multi-threaded environment.
bool	mStopAfterCycle
	If true, then stop at the end of the cycle. Used in multi-threaded environment.
ObjRegistry< RefinableObj >	mRefinedObjList
	The refined objects.
ObjRegistry< RefinableObj >	mRecursiveRefinedObjList
	The refined objects, recursively including all sub-objects.
RefObjOpt	mXMLAutoSave
	Periodic save of complete environment as an xml file.
REAL	mLastOptimTime
	The time elapsed after the last optimization, in seconds.
MainTracker	mMainTracker
	MainTracker object to track the evolution of cost functions, likelihood, and individual parameters.

Detailed Description

Base object for Optimization methods.

This is an abstract base class, derived for Monte-Cralo type algorithms (Simulated Annealing & Parallel Tempering), and hopefully soon for Genetic Algorithms.

Remarks:: Instead of keeping a copy of the list of parameters here, maybe it would be better to delegate all parameter handling to the refined objects (they would also have to keep in memory the saved parameter sets, so that could be difficult to administrate...).

Member Function Documentation

void ObjCryst::OptimizationObj::BuildRecursiveRefObjList ( )

protected

(Re)build OptimizationObj::mRecursiveRefinedObjList, if an object has been added or modified.

If no object has been added and no sub-object has been added/removed, then nothing is done.

RefinableObj& ObjCryst::OptimizationObj::GetFullRefinableObj ( const bool rebuild = true )

Get the RefinableObj with all the parameters from all refined objects.

If rebuild=true, prepare again the list of objects/parameters.

virtual REAL ObjCryst::OptimizationObj::GetLogLikelihood ( ) const

virtual

The optimized (minimized, actually) function.

This function is the weighted sum of the chosen Cost Functions for the refined objects.

virtual void ObjCryst::OptimizationObj::MultiRunOptimize	(	long &	nbCycle,
		long &	nbSteps,
		const bool	silent = `false`,
		const REAL	finalcost = `0`,
		const REAL	maxTime = `-1`
	)

pure virtual

Launch optimization for multiple runs of N steps.

Parameters:

nbCycle,:	the number of runs (cycles) to perform. The structure is randomized at the beginning of each cycle. If nbCycle==-1, this will run indefinitely. The nbCycle parameter is decreased after each run.
nbSteps,:	the number of steps to go. This number is modified (decreases!) as the refinement goes on.
silent	: if true, absolutely no message should be printed (except debugging)
finalcost,:	the optimization will stop if overall cost fallse below this value
maxTime,:	the optimization will stop after the given number of seconds has been spent optimizing (ignored if <0).

Implemented in ObjCryst::MonteCarloObj, and ObjCryst::SimplexObj.

virtual void ObjCryst::OptimizationObj::Optimize	(	long &	nbSteps,
		const bool	silent = `false`,
		const REAL	finalcost = `0`,
		const REAL	maxTime = `-1`
	)

pure virtual

Launch optimization (a single run) for N steps.

Parameters:

nbSteps,:	the number of steps to go. This number is modified (decreases!) as the refinement goes on.
silent	: if true, absolutely no message should be printed (except debugging)
finalcost,:	the optimization will stop if overall cost fallse below this value
maxTime,:	the optimization will stop after the given number of seconds has been spent optimizing (ignored if <0).

Implemented in ObjCryst::MonteCarloObj, and ObjCryst::SimplexObj.

void ObjCryst::OptimizationObj::PrepareRefParList ( )

protected

Prepare mRefParList for the refinement

virtual void ObjCryst::OptimizationObj::RandomizeStartingConfig ( )

virtual

Randomize starting configuration.

Only affects limited and periodic parameters.

virtual void ObjCryst::OptimizationObj::UpdateDisplay ( )

protectedvirtual

Update Display (if any display is available), when a new 'relevant' configuration is reached.

This calls all RefinableObj::UpdateDisplay()

virtual void ObjCryst::OptimizationObj::XMLInput	(	istream &	is,
		const XMLCrystTag &	tag
	)

pure virtual

Input in XML format from a stream, restoring the set of refined objects and the associated cost functions.

Note that the corresponding objects must have been loaded in memory before, else shit happens.

Implemented in ObjCryst::MonteCarloObj, and ObjCryst::SimplexObj.

virtual void ObjCryst::OptimizationObj::XMLOutput	(	ostream &	os,
		int	indent = `0`
	)		const

pure virtual

Output a description of the object in XML format to a stream.

This saves the list of refined object and the cost functions, as well as options for the refinement. The refined objects are not saved, so this must be done somewhere else (they must be reloaded before this object).

Implemented in ObjCryst::MonteCarloObj, and ObjCryst::SimplexObj.

Member Data Documentation

MainTracker ObjCryst::OptimizationObj::mMainTracker

protected

MainTracker object to track the evolution of cost functions, likelihood, and individual parameters.

ObjRegistry<RefinableObj> ObjCryst::OptimizationObj::mRecursiveRefinedObjList

mutableprotected

The refined objects, recursively including all sub-objects.

This is mutable, since it is a function of mRefinedObjList only.

RefinableObj ObjCryst::OptimizationObj::mRefParList

mutableprotected

The refinable par list used during refinement.

Only a condensed version of all objects. This is useful to keep an history of modifications, and to restore previous values.

Remarks:: maybe this should be completely delegated to the refined objetcs.

std::vector<pair<long,REAL> > ObjCryst::OptimizationObj::mvSavedParamSet

protected

List of saved parameter sets.

This is used to save possible solutions during the optimization, so that the user can check them afterwards.

The first member of each pair is the index of the parameter set, and the second is the overall cost for that set.

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

GlobalOptimObj.h

Classes

Public Member Functions

Protected Member Functions

Protected Attributes

Detailed Description

Member Function Documentation

Member Data Documentation