Class to deal with training of Feed-Forward 2 layers Neural Network using IpOpt. More...

#include <neuralNetworks.h>

Inheritance diagram for iCub::optimization::ff2LayNNTrain:

Public Member Functions
	ff2LayNNTrain ()
	Default constructor.
void	setBounds (const yarp::os::Property &bounds)
	Allow specifying the bounds for training network's parameters.
virtual bool	train (const unsigned int numHiddenNodes, const std::deque< yarp::sig::Vector > &in, const std::deque< yarp::sig::Vector > &out, std::deque< yarp::sig::Vector > &pred, double &error)
	Train the network through optimization.
virtual bool	retrain (const std::deque< yarp::sig::Vector > &in, const std::deque< yarp::sig::Vector > &out, std::deque< yarp::sig::Vector > &pred, double &error)
	Retrain the network through optimization.
virtual	~ff2LayNNTrain ()
	Default destructor.
Public Member Functions inherited from iCub::ctrl::ff2LayNN
	ff2LayNN ()
	Create an empty network.
	ff2LayNN (const yarp::os::Property &options)
	Create and configure the network.
virtual bool	configure (const yarp::os::Property &options)
	Configure/reconfigure the network.
virtual bool	isValid () const
	Return the internal status after a configuration.
virtual yarp::sig::Vector	predict (const yarp::sig::Vector &x) const
	Predict the output given a certain input to the network.
virtual bool	getStructure (yarp::os::Property &options) const
	Retrieve the network structure as a Property object.
virtual bool	printStructure (std::ostream &stream=std::cout) const
	Dump tadily the network structure on the stream.
std::deque< yarp::sig::Vector > &	get_IW ()
	Retrieve first layer weights.
std::deque< yarp::sig::Vector > &	get_LW ()
	Retrieve second layer weights.
yarp::sig::Vector &	get_b1 ()
	Retrieve first layer bias.
yarp::sig::Vector &	get_b2 ()
	Retrieve second layer bias.
virtual yarp::sig::Vector	scaleInputToNetFormat (const yarp::sig::Vector &x) const
	Scale input to be used with the network.
virtual yarp::sig::Vector	scaleInputFromNetFormat (const yarp::sig::Vector &x) const
	Scale back input from the network's format.
virtual yarp::sig::Vector	scaleOutputToNetFormat (const yarp::sig::Vector &x) const
	Scale output to be used with the network.
virtual yarp::sig::Vector	scaleOutputFromNetFormat (const yarp::sig::Vector &x) const
	Scale back output from the network's format.
virtual yarp::sig::Vector	hiddenLayerFcn (const yarp::sig::Vector &x) const =0
	Hidden Layer Function.
virtual yarp::sig::Vector	outputLayerFcn (const yarp::sig::Vector &x) const =0
	Output Layer Function.
virtual yarp::sig::Vector	hiddenLayerGrad (const yarp::sig::Vector &x) const =0
	Gradient of the Hidden Layer Function.
virtual yarp::sig::Vector	outputLayerGrad (const yarp::sig::Vector &x) const =0
	Gradient of the Output Layer Function.

Protected Attributes
yarp::os::Property	bounds
void *	App
Protected Attributes inherited from iCub::ctrl::ff2LayNN
std::deque< yarp::sig::Vector >	IW
yarp::sig::Vector	b1
std::deque< yarp::sig::Vector >	LW
yarp::sig::Vector	b2
std::deque< minmax >	inMinMaxX
std::deque< minmax >	inMinMaxY
std::deque< minmax >	outMinMaxX
std::deque< minmax >	outMinMaxY
yarp::sig::Vector	inMinX
yarp::sig::Vector	inMinY
yarp::sig::Vector	inRatio
yarp::sig::Vector	outMinX
yarp::sig::Vector	outMinY
yarp::sig::Vector	outRatio
bool	configured

Additional Inherited Members
Protected Member Functions inherited from iCub::ctrl::ff2LayNN
void	prepare ()
void	setItem (yarp::os::Property &options, const std::string &tag, const yarp::sig::Vector &item) const
bool	getItem (const yarp::os::Property &options, const std::string &tag, yarp::sig::Vector &item) const

Detailed Description

Class to deal with training of Feed-Forward 2 layers Neural Network using IpOpt.

Definition at line 53 of file neuralNetworks.h.

Constructor & Destructor Documentation

◆ ff2LayNNTrain()

ff2LayNNTrain::ff2LayNNTrain ( )

Default constructor.

Definition at line 294 of file neuralNetworks.cpp.

◆ ~ff2LayNNTrain()

ff2LayNNTrain::~ff2LayNNTrain ( )

virtual

Default destructor.

Definition at line 418 of file neuralNetworks.cpp.

Member Function Documentation

◆ retrain()

bool ff2LayNNTrain::retrain	(	const std::deque< yarp::sig::Vector > &	in,
		const std::deque< yarp::sig::Vector > &	out,
		std::deque< yarp::sig::Vector > &	pred,
		double &	error )

virtual

Retrain the network through optimization.

Parameters

numHiddenNodes	is the number of hidden nodes.
in	the list of input vector.
out	the list of output vector.
pred	the returned list of predicted output.
error	returns the prediction error.

Returns: true/false on success/fail.

Definition at line 403 of file neuralNetworks.cpp.

◆ setBounds()

void ff2LayNNTrain::setBounds ( const yarp::os::Property & bounds )

Allow specifying the bounds for training network's parameters.

Parameters

bounds a property-like object containing bounds in the form: ("tag" (<min> <max>)), where tag is a string referring to network parameters: e.g. IW, LW, b1, b2 ...

Definition at line 311 of file neuralNetworks.cpp.

◆ train()

bool ff2LayNNTrain::train	(	const unsigned int	numHiddenNodes,
		const std::deque< yarp::sig::Vector > &	in,
		const std::deque< yarp::sig::Vector > &	out,
		std::deque< yarp::sig::Vector > &	pred,
		double &	error )

virtual

Train the network through optimization.

Parameters

numHiddenNodes	is the number of hidden nodes.
in	the list of input vector.
out	the list of output vector.
pred	the returned list of predicted output.
error	returns the prediction error.

Returns: true/false on success/fail.

Definition at line 318 of file neuralNetworks.cpp.

Member Data Documentation

◆ App

void* iCub::optimization::ff2LayNNTrain::App

protected

Definition at line 57 of file neuralNetworks.h.

◆ bounds

yarp::os::Property iCub::optimization::ff2LayNNTrain::bounds

protected

Definition at line 56 of file neuralNetworks.h.

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

icub-main/src/libraries/optimization/include/iCub/optimization/neuralNetworks.h
icub-main/src/libraries/optimization/src/neuralNetworks.cpp

Public Member Functions

Protected Attributes

Additional Inherited Members

Detailed Description

Constructor & Destructor Documentation

◆ ff2LayNNTrain()

◆ ~ff2LayNNTrain()

Member Function Documentation

◆ retrain()

◆ setBounds()

◆ train()

Member Data Documentation

◆ App

◆ bounds