Artificial Neural Networks - Multi-Layer Perceptrons. More...

#include "ml.hpp"

Inheritance diagram for cv::ml::ANN_MLP:

Collaboration diagram for cv::ml::ANN_MLP:

Public Types
enum	ActivationFunctions { IDENTITY = 0, SIGMOID_SYM = 1, GAUSSIAN = 2 }
	possible activation functions More...

enum	Flags { UPDATE_MODEL = 1, RAW_OUTPUT =1, COMPRESSED_INPUT =2, PREPROCESSED_INPUT =4 }
	Predict options. More...

enum	TrainFlags { UPDATE_WEIGHTS = 1, NO_INPUT_SCALE = 2, NO_OUTPUT_SCALE = 4 }
	Train options. More...

enum	TrainingMethods { BACKPROP =0, RPROP =1 }
	Available training methods. More...

Public Member Functions
virtual float	calcError (const Ptr< TrainData > &data, bool test, OutputArray resp) const
	Computes error on the training or test dataset. More...

virtual void	clear ()
	Clears the algorithm state. More...

virtual bool	empty () const
	Returns true if the Algorithm is empty (e.g. More...

virtual double	getBackpropMomentumScale () const =0
	BPROP: Strength of the momentum term (the difference between weights on the 2 previous iterations). More...

virtual double	getBackpropWeightScale () const =0
	BPROP: Strength of the weight gradient term. More...

virtual String	getDefaultName () const
	Returns the algorithm string identifier. More...

virtual cv::Mat	getLayerSizes () const =0
	Integer vector specifying the number of neurons in each layer including the input and output layers. More...

virtual double	getRpropDW0 () const =0
	RPROP: Initial value \(\Delta_0\) of update-values \(\Delta_{ij}\). More...

virtual double	getRpropDWMax () const =0
	RPROP: Update-values upper limit \(\Delta_{max}\). More...

virtual double	getRpropDWMin () const =0
	RPROP: Update-values lower limit \(\Delta_{min}\). More...

virtual double	getRpropDWMinus () const =0
	RPROP: Decrease factor \(\eta^-\). More...

virtual double	getRpropDWPlus () const =0
	RPROP: Increase factor \(\eta^+\). More...

virtual TermCriteria	getTermCriteria () const =0
	Termination criteria of the training algorithm. More...

virtual int	getTrainMethod () const =0
	Returns current training method. More...

virtual int	getVarCount () const =0
	Returns the number of variables in training samples. More...

virtual Mat	getWeights (int layerIdx) const =0

virtual bool	isClassifier () const =0
	Returns true if the model is classifier. More...

virtual bool	isTrained () const =0
	Returns true if the model is trained. More...

virtual float	predict (InputArray samples, OutputArray results=noArray(), int flags=0) const =0
	Predicts response(s) for the provided sample(s) More...

virtual void	read (const FileNode &fn)
	Reads algorithm parameters from a file storage. More...

virtual void	save (const String &filename) const
	Saves the algorithm to a file. More...

virtual void	setActivationFunction (int type, double param1=0, double param2=0)=0
	Initialize the activation function for each neuron. More...

virtual void	setBackpropMomentumScale (double val)=0
	BPROP: Strength of the momentum term (the difference between weights on the 2 previous iterations). More...

virtual void	setBackpropWeightScale (double val)=0
	BPROP: Strength of the weight gradient term. More...

virtual void	setLayerSizes (InputArray _layer_sizes)=0
	Integer vector specifying the number of neurons in each layer including the input and output layers. More...

virtual void	setRpropDW0 (double val)=0
	RPROP: Initial value \(\Delta_0\) of update-values \(\Delta_{ij}\). More...

virtual void	setRpropDWMax (double val)=0
	RPROP: Update-values upper limit \(\Delta_{max}\). More...

virtual void	setRpropDWMin (double val)=0
	RPROP: Update-values lower limit \(\Delta_{min}\). More...

virtual void	setRpropDWMinus (double val)=0
	RPROP: Decrease factor \(\eta^-\). More...

virtual void	setRpropDWPlus (double val)=0
	RPROP: Increase factor \(\eta^+\). More...

virtual void	setTermCriteria (TermCriteria val)=0
	Termination criteria of the training algorithm. More...

virtual void	setTrainMethod (int method, double param1=0, double param2=0)=0
	Sets training method and common parameters. More...

virtual bool	train (const Ptr< TrainData > &trainData, int flags=0)
	Trains the statistical model. More...

virtual bool	train (InputArray samples, int layout, InputArray responses)
	Trains the statistical model. More...

virtual void	write (FileStorage &fs) const
	Stores algorithm parameters in a file storage. More...

Static Public Member Functions
static Ptr< ANN_MLP >	create ()
	Creates empty model. More...

static Ptr< ANN_MLP >	load (const String &filepath)
	Loads and creates a serialized ANN from a file. More...

template<typename _Tp >
static Ptr< _Tp >	load (const String &filename, const String &objname=String())
	Loads algorithm from the file. More...

template<typename _Tp >
static Ptr< _Tp >	loadFromString (const String &strModel, const String &objname=String())
	Loads algorithm from a String. More...

template<typename _Tp >
static Ptr< _Tp >	read (const FileNode &fn)
	Reads algorithm from the file node. More...

template<typename _Tp >
static Ptr< _Tp >	train (const Ptr< TrainData > &data, int flags=0)
	Create and train model with default parameters. More...

Protected Member Functions
void	writeFormat (FileStorage &fs) const

Detailed Description

Artificial Neural Networks - Multi-Layer Perceptrons.

Unlike many other models in ML that are constructed and trained at once, in the MLP model these steps are separated. First, a network with the specified topology is created using the non-default constructor or the method ANN_MLP::create. All the weights are set to zeros. Then, the network is trained using a set of input and output vectors. The training procedure can be repeated more than once, that is, the weights can be adjusted based on the new training data.

Additional flags for StatModel::train are available: ANN_MLP::TrainFlags.

See also: Neural Networks

Member Enumeration Documentation

enum cv::ml::ANN_MLP::ActivationFunctions

possible activation functions

Enumerator

IDENTITY

Identity function: \(f(x)=x\).

SIGMOID_SYM

Symmetrical sigmoid: \(f(x)=\beta*(1-e^{-\alpha x})/(1+e^{-\alpha x}\).

Note: If you are using the default sigmoid activation function with the default parameter values fparam1=0 and fparam2=0 then the function used is y = 1.7159*tanh(2/3 * x), so the output will range from [-1.7159, 1.7159], instead of [0,1].

GAUSSIAN

Gaussian function: \(f(x)=\beta e^{-\alpha x*x}\).

enum cv::ml::StatModel::Flags

inherited

Predict options.

Enumerator
UPDATE_MODEL
RAW_OUTPUT	makes the method return the raw results (the sum), not the class label
COMPRESSED_INPUT
PREPROCESSED_INPUT

enum cv::ml::ANN_MLP::TrainFlags

Train options.

Enumerator

UPDATE_WEIGHTS

Update the network weights, rather than compute them from scratch.

In the latter case the weights are initialized using the Nguyen-Widrow algorithm.

NO_INPUT_SCALE

Do not normalize the input vectors.

If this flag is not set, the training algorithm normalizes each input feature independently, shifting its mean value to 0 and making the standard deviation equal to 1. If the network is assumed to be updated frequently, the new training data could be much different from original one. In this case, you should take care of proper normalization.

NO_OUTPUT_SCALE

Do not normalize the output vectors.

If the flag is not set, the training algorithm normalizes each output feature independently, by transforming it to the certain range depending on the used activation function.

enum cv::ml::ANN_MLP::TrainingMethods

Available training methods.

Enumerator

BACKPROP

The back-propagation algorithm.

RPROP

The RPROP algorithm.

See [68] for details.

Member Function Documentation

virtual float cv::ml::StatModel::calcError	(	const Ptr< TrainData > &	data,
		bool	test,
		OutputArray	resp
	)		const

virtualinherited

Computes error on the training or test dataset.

Parameters

data	the training data
test	if true, the error is computed over the test subset of the data, otherwise it's computed over the training subset of the data. Please note that if you loaded a completely different dataset to evaluate already trained classifier, you will probably want not to set the test subset at all with TrainData::setTrainTestSplitRatio and specify test=false, so that the error is computed for the whole new set. Yes, this sounds a bit confusing.
resp	the optional output responses.

The method uses StatModel::predict to compute the error. For regression models the error is computed as RMS, for classifiers - as a percent of missclassified samples (0%-100%).

virtual void cv::Algorithm::clear ( )

inlinevirtualinherited

Clears the algorithm state.

Reimplemented in cv::FlannBasedMatcher, cv::DescriptorMatcher, and cv::cuda::DescriptorMatcher.

static Ptr<ANN_MLP> cv::ml::ANN_MLP::create ( )

static

Creates empty model.

Use StatModel::train to train the model, Algorithm::load<ANN_MLP>(filename) to load the pre-trained model. Note that the train method has optional flags: ANN_MLP::TrainFlags.

virtual bool cv::ml::StatModel::empty ( ) const

virtualinherited

Returns true if the Algorithm is empty (e.g.

in the very beginning or after unsuccessful read

Reimplemented from cv::Algorithm.

virtual double cv::ml::ANN_MLP::getBackpropMomentumScale ( ) const

pure virtual

BPROP: Strength of the momentum term (the difference between weights on the 2 previous iterations).

This parameter provides some inertia to smooth the random fluctuations of the weights. It can vary from 0 (the feature is disabled) to 1 and beyond. The value 0.1 or so is good enough. Default value is 0.1.

See also: setBackpropMomentumScale

virtual double cv::ml::ANN_MLP::getBackpropWeightScale ( ) const

pure virtual

BPROP: Strength of the weight gradient term.

The recommended value is about 0.1. Default value is 0.1.

See also: setBackpropWeightScale

virtual String cv::Algorithm::getDefaultName ( ) const

virtualinherited

Returns the algorithm string identifier.

This string is used as top level xml/yml node tag when the object is saved to a file or string.

virtual cv::Mat cv::ml::ANN_MLP::getLayerSizes ( ) const

pure virtual

Integer vector specifying the number of neurons in each layer including the input and output layers.

The very first element specifies the number of elements in the input layer. The last element - number of elements in the output layer.

See also: setLayerSizes

virtual double cv::ml::ANN_MLP::getRpropDW0 ( ) const

pure virtual

RPROP: Initial value \(\Delta_0\) of update-values \(\Delta_{ij}\).

Default value is 0.1.

See also: setRpropDW0

virtual double cv::ml::ANN_MLP::getRpropDWMax ( ) const

pure virtual

RPROP: Update-values upper limit \(\Delta_{max}\).

It must be >1. Default value is 50.

See also: setRpropDWMax

virtual double cv::ml::ANN_MLP::getRpropDWMin ( ) const

pure virtual

RPROP: Update-values lower limit \(\Delta_{min}\).

It must be positive. Default value is FLT_EPSILON.

See also: setRpropDWMin

virtual double cv::ml::ANN_MLP::getRpropDWMinus ( ) const

pure virtual

RPROP: Decrease factor \(\eta^-\).

It must be <1. Default value is 0.5.

See also: setRpropDWMinus

virtual double cv::ml::ANN_MLP::getRpropDWPlus ( ) const

pure virtual

RPROP: Increase factor \(\eta^+\).

It must be >1. Default value is 1.2.

See also: setRpropDWPlus

virtual TermCriteria cv::ml::ANN_MLP::getTermCriteria ( ) const

pure virtual

Termination criteria of the training algorithm.

You can specify the maximum number of iterations (maxCount) and/or how much the error could change between the iterations to make the algorithm continue (epsilon). Default value is TermCriteria(TermCriteria::MAX_ITER + TermCriteria::EPS, 1000, 0.01).

See also: setTermCriteria

virtual int cv::ml::ANN_MLP::getTrainMethod ( ) const

pure virtual

Returns current training method.

virtual int cv::ml::StatModel::getVarCount ( ) const

pure virtualinherited

Returns the number of variables in training samples.

virtual Mat cv::ml::ANN_MLP::getWeights ( int layerIdx ) const

pure virtual

virtual bool cv::ml::StatModel::isClassifier ( ) const

pure virtualinherited

Returns true if the model is classifier.

virtual bool cv::ml::StatModel::isTrained ( ) const

pure virtualinherited

Returns true if the model is trained.

static Ptr<ANN_MLP> cv::ml::ANN_MLP::load ( const String & filepath )

static

Loads and creates a serialized ANN from a file.

Use ANN::save to serialize and store an ANN to disk. Load the ANN from this file again, by calling this function with the path to the file.

Parameters

filepath path to serialized ANN

template<typename _Tp >

static Ptr<_Tp> cv::Algorithm::load	(	const String &	filename,
		const String &	objname = `String()`
	)

inlinestaticinherited

Loads algorithm from the file.

Parameters

filename	Name of the file to read.
objname	The optional name of the node to read (if empty, the first top-level node will be used)

This is static template method of Algorithm. It's usage is following (in the case of SVM):

Ptr<SVM> svm = Algorithm::load<SVM>("my_svm_model.xml");

In order to make this method work, the derived class must overwrite Algorithm::read(const FileNode& fn).

References cv::Ptr< T >::empty(), cv::FileNode::empty(), cv::FileStorage::getFirstTopLevelNode(), and cv::FileStorage::READ.

Here is the call graph for this function:

template<typename _Tp >

static Ptr<_Tp> cv::Algorithm::loadFromString	(	const String &	strModel,
		const String &	objname = `String()`
	)

inlinestaticinherited

Loads algorithm from a String.

Parameters

strModel	The string variable containing the model you want to load.
objname	The optional name of the node to read (if empty, the first top-level node will be used)

This is static template method of Algorithm. It's usage is following (in the case of SVM):

Ptr<SVM> svm = Algorithm::loadFromString<SVM>(myStringModel);

References CV_WRAP, cv::Ptr< T >::empty(), cv::FileNode::empty(), cv::FileStorage::getFirstTopLevelNode(), cv::FileStorage::MEMORY, and cv::FileStorage::READ.

Here is the call graph for this function:

virtual float cv::ml::StatModel::predict	(	InputArray	samples,
		OutputArray	results = `noArray()`,
		int	flags = `0`
	)		const

pure virtualinherited

Predicts response(s) for the provided sample(s)

Parameters

samples	The input samples, floating-point matrix
results	The optional output matrix of results.
flags	The optional flags, model-dependent. See cv::ml::StatModel::Flags.

Implemented in cv::ml::LogisticRegression, and cv::ml::EM.

virtual void cv::Algorithm::read ( const FileNode & fn )

inlinevirtualinherited

Reads algorithm parameters from a file storage.

Reimplemented in cv::FlannBasedMatcher, cv::DescriptorMatcher, and cv::Feature2D.

template<typename _Tp >

static Ptr<_Tp> cv::Algorithm::read ( const FileNode & fn )

inlinestaticinherited

Reads algorithm from the file node.

This is static template method of Algorithm. It's usage is following (in the case of SVM):

cv::FileStorage fsRead("example.xml", FileStorage::READ);

Ptr<SVM> svm = Algorithm::read<SVM>(fsRead.root());

In order to make this method work, the derived class must overwrite Algorithm::read(const FileNode& fn) and also have static create() method without parameters (or with all the optional parameters)

References cv::Ptr< T >::empty().

Here is the call graph for this function:

virtual void cv::Algorithm::save ( const String & filename ) const

virtualinherited

Saves the algorithm to a file.

In order to make this method work, the derived class must implement Algorithm::write(FileStorage& fs).

virtual void cv::ml::ANN_MLP::setActivationFunction	(	int	type,
		double	param1 = `0`,
		double	param2 = `0`
	)

pure virtual

Initialize the activation function for each neuron.

Currently the default and the only fully supported activation function is ANN_MLP::SIGMOID_SYM.

Parameters

type	The type of activation function. See ANN_MLP::ActivationFunctions.
param1	The first parameter of the activation function, \(\alpha\). Default value is 0.
param2	The second parameter of the activation function, \(\beta\). Default value is 0.

virtual void cv::ml::ANN_MLP::setBackpropMomentumScale ( double val )

pure virtual

BPROP: Strength of the momentum term (the difference between weights on the 2 previous iterations).

See also: getBackpropMomentumScale

virtual void cv::ml::ANN_MLP::setBackpropWeightScale ( double val )

pure virtual

BPROP: Strength of the weight gradient term.

See also: getBackpropWeightScale

virtual void cv::ml::ANN_MLP::setLayerSizes ( InputArray _layer_sizes )

pure virtual

Integer vector specifying the number of neurons in each layer including the input and output layers.

The very first element specifies the number of elements in the input layer. The last element - number of elements in the output layer. Default value is empty Mat.

See also: getLayerSizes

virtual void cv::ml::ANN_MLP::setRpropDW0 ( double val )

pure virtual

RPROP: Initial value \(\Delta_0\) of update-values \(\Delta_{ij}\).

See also: getRpropDW0

virtual void cv::ml::ANN_MLP::setRpropDWMax ( double val )

pure virtual

RPROP: Update-values upper limit \(\Delta_{max}\).

See also: getRpropDWMax

virtual void cv::ml::ANN_MLP::setRpropDWMin ( double val )

pure virtual

RPROP: Update-values lower limit \(\Delta_{min}\).

See also: getRpropDWMin

virtual void cv::ml::ANN_MLP::setRpropDWMinus ( double val )

pure virtual

RPROP: Decrease factor \(\eta^-\).

See also: getRpropDWMinus

virtual void cv::ml::ANN_MLP::setRpropDWPlus ( double val )

pure virtual

RPROP: Increase factor \(\eta^+\).

See also: getRpropDWPlus

virtual void cv::ml::ANN_MLP::setTermCriteria ( TermCriteria val )

pure virtual

Termination criteria of the training algorithm.

See also: getTermCriteria

virtual void cv::ml::ANN_MLP::setTrainMethod	(	int	method,
		double	param1 = `0`,
		double	param2 = `0`
	)

pure virtual

Sets training method and common parameters.

Parameters

method	Default value is ANN_MLP::RPROP. See ANN_MLP::TrainingMethods.
param1	passed to setRpropDW0 for ANN_MLP::RPROP and to setBackpropWeightScale for ANN_MLP::BACKPROP
param2	passed to setRpropDWMin for ANN_MLP::RPROP and to setBackpropMomentumScale for ANN_MLP::BACKPROP.

virtual bool cv::ml::StatModel::train	(	const Ptr< TrainData > &	trainData,
		int	flags = `0`
	)

virtualinherited

Trains the statistical model.

Parameters

trainData	training data that can be loaded from file using TrainData::loadFromCSV or created with TrainData::create.
flags	optional flags, depending on the model. Some of the models can be updated with the new training samples, not completely overwritten (such as NormalBayesClassifier or ANN_MLP).

virtual bool cv::ml::StatModel::train	(	InputArray	samples,
		int	layout,
		InputArray	responses
	)

virtualinherited

Trains the statistical model.

Parameters

samples	training samples
layout	See ml::SampleTypes.
responses	vector of responses associated with the training samples.

template<typename _Tp >

static Ptr<_Tp> cv::ml::StatModel::train	(	const Ptr< TrainData > &	data,
		int	flags = `0`
	)

inlinestaticinherited

Create and train model with default parameters.

The class must implement static create() method with no parameters or with all default parameter values

References cv::Ptr< T >::empty().

Here is the call graph for this function:

virtual void cv::Algorithm::write ( FileStorage & fs ) const

inlinevirtualinherited

Stores algorithm parameters in a file storage.

Reimplemented in cv::FlannBasedMatcher, cv::DescriptorMatcher, and cv::Feature2D.

void cv::Algorithm::writeFormat ( FileStorage & fs ) const

protectedinherited

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

ml/include/opencv2/ml.hpp

Public Types

Public Member Functions

Static Public Member Functions

Protected Member Functions

Detailed Description

Member Enumeration Documentation

Member Function Documentation