Reinforcement Learning Toolbox 2.0
last updated:

Documentation

	Manual
	Tutorial
	Class Reference

Master Thesis

Examples

CFeatureList Class Reference

Class for storing features as sparse array. More...

#include <cfeaturefunction.h>

List of all members.


Public Types
typedef CFeatureList::iterator	iterator
typedef CFeatureList::reverse_iterator	reverse_iterator
Public Member Functions
	CFeatureList (int initMemSize=0, bool isSorted=false, bool sortAbs=false)
	Create a new feature list.
	~CFeatureList ()
CFeatureList::iterator	getFeaturePos (unsigned int feature)
	Returns iterator pointing on the given feature (if its in the list, other wise end will be returned).
void	add (CFeature *feature)
	Add a CFeature object to the list.
void	add (CFeatureList *featureList, double factor=1.0)
	Add all features of the given list to the current feature list. Thre feature factors of the given list are multiplied by "factor".
void	set (int feature, double factor)
	Set the given feature to the given factor.
void	multFactor (double factor)
	Multiply all feature factors with the given factor.
double	multFeatureList (CFeatureList *featureList)
	Multiply all feature of the given list with the features of the current list.
void	addIndexOffset (int Offset)
	Add to all features the given index offset.
void	update (int feature, double factor)
	Add the given feature factor to the given feature.
double	getFeatureFactor (int featureIndex)
	Returns the feature factor of the given feature.
CFeature *	getFeature (int featureIndex)
	Returns the whole CFeature object with the given factor.
void	remove (CFeature *feature)
	removes the given feature from the list
void	remove (int feature)
void	clear ()
	clears the feature list
void	clearAndDelete ()
	clears the feature list and deletes all CFeature objects
void	saveASCII (FILE *stream)
	save the feature list to a ascii stream
void	loadASCII (FILE *stream)
	load the feature list to a ascii stream
void	normalize ()
	normalizes the feature list (sum of all factors = 1)
double	getLength ()
	returns the euklidean length of the feature space
CFeatureList::iterator	begin ()
CFeatureList::iterator	end ()
CFeatureList::reverse_iterator	rbegin ()
CFeatureList::reverse_iterator	rend ()
int	size ()
Protected Member Functions
CFeature *	getFreeFeatureResource ()
	Internal function, used for resource management.
void	sortFeature (CFeature *feature)
	insert the given feature in the correct position.
Protected Attributes
std::list< CFeature * > *	freeResources
std::list< CFeature * > *	allResources
std::map< int, CFeature * > *	featureMap
bool	isSorted
bool	sortAbs

Detailed Description

Class for storing features as sparse array.

The class maintains represents a list of features and is often used for gradient calculatation. In a feature list, not all features have to be included, the features which are not in the list have automatically the value 0.0. Every feature in the list is stored as a CFeature object, so the index and the feature factor are stored. The CFeatureList class also contains methods for directly setting (set), updating (adding a factor, update) or getting (getFeatureFactor) feature factors, without the need of CFeature objects. CFeatureList implements all functionality of the std::list class, so iterators can be used in the same way. The class provides its own resource management, so it always uses its own CFeature objects, so you don't have to care about where the CFeature objects come from or where they will go (cleaning up). If the feature list gets too small, it automatically allocates new resources. If the list gets smaller again, it always holds the allocated resources (even if you call clear) because its likely that the resources will be needed again. The allocated CFeature resources get deleted only when the feature list itself is deleted or clearAndDelete is called. For performance reasons the feature list also contains a map of CFeature object, with its feature index as search index. So the access to a particular feature is very quick. All features that are in the list are always in the map too. You have also the possibility to add another feature list to the current list (feature factors are added, non existant features in one list count as feature factor 0, or to multiply 2 feature lists (needed for gradient multiplication). The multiplied feature list naturally contains only the features which are in both components of the product. Feature lists can also be sorted by the feature factor. This is done for example in the case of etraces. If you need a sorted feature list, set the isSortet flag of the constructor to true. The first feature in the list will then always be the one with the highest factor.

Member Typedef Documentation

typedef CFeatureList::iterator CFeatureList::iterator

typedef CFeatureList::reverse_iterator CFeatureList::reverse_iterator

Constructor & Destructor Documentation

CFeatureList::CFeatureList	(	int	initMemSize = `0`,
		bool	isSorted = `false`,
		bool	sortAbs = `false`
	)

Create a new feature list.

The parameter initMemSize specifies the number of CFeatures object that are allocated in the beginning. The second flag specifies wether the list is sorted.

CFeatureList::~CFeatureList

(

)

Member Function Documentation

void CFeatureList::add	(	CFeatureList *	featureList,
		double	factor = `1.0`
	)

Add all features of the given list to the current feature list. Thre feature factors of the given list are multiplied by "factor".

void CFeatureList::add

(

CFeature *

feature

)

Add a CFeature object to the list.

The value of the feature is added to the feature in the list or, if the feature isn't already in the list, a new feature is added to the list. Even though a CFeature object is given, the feature list uses its own CFeature objects.

void CFeatureList::addIndexOffset

(

int

Offset

)

Add to all features the given index offset.

CFeatureList::iterator CFeatureList::begin

(

)

void CFeatureList::clear

(

)

clears the feature list

The allocated CFeature objects remain in memory, they are just removed from the list and put in the freeResources list.

void CFeatureList::clearAndDelete

(

)

clears the feature list and deletes all CFeature objects

CFeatureList::iterator CFeatureList::end

(

)

CFeature* CFeatureList::getFeature

(

int

featureIndex

)

Returns the whole CFeature object with the given factor.

If the specified feature isn't in the list, NULL is returned. Don't change any values or delete that CFeature object !!!

double CFeatureList::getFeatureFactor

(

int

featureIndex

)

Returns the feature factor of the given feature.

CFeatureList::iterator CFeatureList::getFeaturePos

(

unsigned int

feature

)

Returns iterator pointing on the given feature (if its in the list, other wise end will be returned).

CFeature* CFeatureList::getFreeFeatureResource ( ) [protected]

Internal function, used for resource management.

Returns an unused or a new CFeature object.

double CFeatureList::getLength

(

)

returns the euklidean length of the feature space

void CFeatureList::loadASCII

(

FILE *

stream

)

load the feature list to a ascii stream

void CFeatureList::multFactor

(

double

factor

)

Multiply all feature factors with the given factor.

double CFeatureList::multFeatureList

(

CFeatureList *

featureList

)

Multiply all feature of the given list with the features of the current list.

This calculation is like the dot product, features that are not in a list have the feature factors 0.0.

void CFeatureList::normalize

(

)

normalizes the feature list (sum of all factors = 1)

CFeatureList::reverse_iterator CFeatureList::rbegin

(

)

void CFeatureList::remove

(

int

feature

)

void CFeatureList::remove

(

CFeature *

feature

)

removes the given feature from the list

CFeatureList::reverse_iterator CFeatureList::rend

(

)

void CFeatureList::saveASCII

(

FILE *

stream

)

save the feature list to a ascii stream

void CFeatureList::set	(	int	feature,
		double	factor
	)

Set the given feature to the given factor.

If the feature is not already in the list a new CFeature object is requested (by the function getFreeFeatureResource.

int CFeatureList::size ( ) [inline]

void CFeatureList::sortFeature ( CFeature * feature ) [protected]

insert the given feature in the correct position.

void CFeatureList::update	(	int	feature,
		double	factor
	)

Add the given feature factor to the given feature.

If the feature is not already in the list a new CFeature object is requested (by the function getFreeFeatureResource) and the feature factor is set according to factor.

Member Data Documentation

std::list<CFeature *>* CFeatureList::allResources [protected]

std::map<int, CFeature *>* CFeatureList::featureMap [protected]

std::list<CFeature *>* CFeatureList::freeResources [protected]

bool CFeatureList::isSorted [protected]

bool CFeatureList::sortAbs [protected]

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

cfeaturefunction.h

CFeatureList Class Reference

Public Types

Public Member Functions

Protected Member Functions

Protected Attributes

Detailed Description

Member Typedef Documentation

Constructor & Destructor Documentation

Member Function Documentation

Member Data Documentation