MultiLayerFFNN Class Reference

multi layer neural network with configurable activation functions More...

#include <multilayerffnn.h>

Inherits FeedForwardNN.

Inherited by Elman.

Inheritance diagram for MultiLayerFFNN:

Inheritance graph
[legend]
Collaboration diagram for MultiLayerFFNN:

Collaboration graph
[legend]
List of all members.

Public Member Functions

 MultiLayerFFNN (double eps, const std::vector< Layer > &layers, bool useBypass=false, bool someInternalParams=true)
virtual ~MultiLayerFFNN ()
virtual void init (unsigned int inputDim, unsigned int outputDim, double unit_map=0.0, RandGen *randGen=0)
 initialisation of the network with the given number of input and output units.
virtual const matrix::Matrix process (const matrix::Matrix &input)
 passive processing of the input
virtual const matrix::Matrix learn (const matrix::Matrix &input, const matrix::Matrix &nom_output, double learnRateFactor=1)
 performs learning and returns the network output before learning (process should be called before)
virtual const matrix::Matrix response (const matrix::Matrix &input) const
 response matrix of neural network at given input
virtual const matrix::Matrix inversion (const matrix::Matrix &input, const matrix::Matrix &xsi) const
 calculates the input shift v to a given output shift xsi via pseudo inversion.
virtual unsigned int getInputDim () const
 returns the number of input neurons
virtual unsigned int getOutputDim () const
 returns the number of output neurons
virtual matrix::Matrix getLayerOutput (unsigned int layer)
 returns activation of the given layer. Layer 0 is the first hidden layer
virtual void damp (double damping)
 damps the weights and the biases by multiplying (1-damping)
virtual unsigned int getLayerNum () const
virtual const Layer & getLayer (unsigned int layer) const
 layers 0 is the first hidden layer
virtual Layer & getLayer (unsigned int layer)
 layers 0 is the first hidden layer
virtual const matrix::Matrix & getWeights (unsigned int to_layer) const
 weight matrix 0 connects input with the first hidden layer
virtual matrix::Matrix & getWeights (unsigned int to_layer)
 weight matrix 0 connects input with the first hidden layer
virtual const matrix::Matrix & getBias (unsigned int of_layer) const
 layers 0 is the first hidden layer
bool store (FILE *f) const
 stores the layer binary into file stream
bool restore (FILE *f)
 restores the layer binary from file stream
bool write (FILE *f) const
 writes the layer ASCII into file stream (not in the storable interface)
virtual iparamkeylist getInternalParamNames () const
 The list of the names of all internal parameters given by getInternalParams().
virtual iparamvallist getInternalParams () const
virtual ilayerlist getStructuralLayers () const
 Specifies which parameter vector forms a structural layer (in terms of a neural network) The ordering is important.
virtual iconnectionlist getStructuralConnections () const
 Specifies which parameter matrix forms a connection between layers (in terms of a neural network) The orderning is not important.
virtual void setSomeInternalParams (bool someInternalParams)
virtual std::vector< ActivationFunction > setActivationFunction (ActivationFunction actfun)
 sets the activation function (and derivative and inversion too) for ALL layers!
virtual void setActivationFunctions (std::vector< ActivationFunction > actfunList)
 sets the activation functions (and derivative and inversion too) for all layers.

Public Attributes

double eps
 learning rate

Protected Attributes

std::vector< Layer > layers
std::vector< matrix::Matrix > weights
std::vector< matrix::Matrix > bias
std::vector< matrix::Matrix > smallids
bool useBypass
matrix::Matrix bypassWeights
bool someInternalParams
matrix::Matrix input
std::vector< matrix::Matrix > ys
std::vector< matrix::Matrix > zs
bool initialised

Detailed Description

multi layer neural network with configurable activation functions

Constructor & Destructor Documentation

MultiLayerFFNN ( double  eps,
const std::vector< Layer > &  layers,
bool  useBypass = false,
bool  someInternalParams = true 
)

Parameters:
eps learning rate
layers Layer description (the input layer is not specified (always linear))
useBypass if true, then a connection from input to output layer is included
someInternalParams if true then only a few parameters are send to plotting

virtual ~MultiLayerFFNN (  )  [inline, virtual]

Member Function Documentation

void damp ( double  damping  )  [virtual]

damps the weights and the biases by multiplying (1-damping)

Implements FeedForwardNN.

Reimplemented in Elman.

virtual const matrix::Matrix& getBias ( unsigned int  of_layer  )  const [inline, virtual]

layers 0 is the first hidden layer

virtual unsigned int getInputDim (  )  const [inline, virtual]

returns the number of input neurons

Implements AbstractModel.

Inspectable::iparamkeylist getInternalParamNames (  )  const [virtual]

The list of the names of all internal parameters given by getInternalParams().

The naming convention is "v[i]" for vectors and "A[i][j]" for matrices, where i, j start at 0.

Returns:
: list of keys

Reimplemented from Inspectable.

Reimplemented in Elman.

Inspectable::iparamvallist getInternalParams (  )  const [virtual]

Returns:
: list of values

Reimplemented from Inspectable.

Reimplemented in Elman.

virtual Layer& getLayer ( unsigned int  layer  )  [inline, virtual]

layers 0 is the first hidden layer

virtual const Layer& getLayer ( unsigned int  layer  )  const [inline, virtual]

layers 0 is the first hidden layer

virtual unsigned int getLayerNum (  )  const [inline, virtual]

virtual matrix::Matrix getLayerOutput ( unsigned int  layer  )  [inline, virtual]

returns activation of the given layer. Layer 0 is the first hidden layer

virtual unsigned int getOutputDim (  )  const [inline, virtual]

returns the number of output neurons

Implements AbstractModel.

Inspectable::iconnectionlist getStructuralConnections (  )  const [virtual]

Specifies which parameter matrix forms a connection between layers (in terms of a neural network) The orderning is not important.

Returns:
: list of layer names with dimension

Reimplemented from Inspectable.

Reimplemented in Elman.

Inspectable::ilayerlist getStructuralLayers (  )  const [virtual]

Specifies which parameter vector forms a structural layer (in terms of a neural network) The ordering is important.

The first entry is the input layer and so on.

Returns:
: list of layer names with dimension

Reimplemented from Inspectable.

Reimplemented in Elman.

virtual matrix::Matrix& getWeights ( unsigned int  to_layer  )  [inline, virtual]

weight matrix 0 connects input with the first hidden layer

virtual const matrix::Matrix& getWeights ( unsigned int  to_layer  )  const [inline, virtual]

weight matrix 0 connects input with the first hidden layer

void init ( unsigned int  inputDim,
unsigned int  outputDim,
double  unit_map = 0.0,
RandGen *  randGen = 0 
) [virtual]

initialisation of the network with the given number of input and output units.

The dimensionality of the ouputlayer is automatically adjusted.

Parameters:
unit_map defines the approximate response of the network after initialisation (if unit_map=1 the weights are unit matrices).
randGen pointer to random generator, if 0 an new one is used

Implements AbstractModel.

Reimplemented in Elman.

const Matrix inversion ( const matrix::Matrix &  input,
const matrix::Matrix &  xsi 
) const [virtual]

calculates the input shift v to a given output shift xsi via pseudo inversion.

\[o+\xi = \psi(i+v)\]

The result is a vector of dimension inputdim. ATTENTION: input is ignored! use process before!

Implements InvertableModel.

const Matrix learn ( const matrix::Matrix &  input,
const matrix::Matrix &  nom_output,
double  learnRateFactor = 1 
) [virtual]

performs learning and returns the network output before learning (process should be called before)

Implements AbstractModel.

Reimplemented in Elman.

const Matrix process ( const matrix::Matrix &  input  )  [virtual]

passive processing of the input

Implements AbstractModel.

Reimplemented in Elman.

const Matrix response ( const matrix::Matrix &  input  )  const [virtual]

response matrix of neural network at given input

\[ J_ij = \frac{\partial y_i}{\partial x_j} \]

\[ J = G_n' W_n G_{n-1}' W_{n-1} ... G_1' W_1 \]

with $W_n$ is the weight matrix of layer n and $ G'$ is a diagonal matrix with $ G'_ii = g'_i $ as values on the diagonal. ATTENTION: input is ignored! use process before!

Implements InvertableModel.

bool restore ( FILE *  f  )  [virtual]

restores the layer binary from file stream

Implements Storeable.

Reimplemented in Elman.

std::vector< ActivationFunction > setActivationFunction ( ActivationFunction  actfun  )  [virtual]

sets the activation function (and derivative and inversion too) for ALL layers!

Parameters:
actfun the activation function to be used
Returns:
the activation functions which where used until now

void setActivationFunctions ( std::vector< ActivationFunction >  actfunList  )  [virtual]

sets the activation functions (and derivative and inversion too) for all layers.

Note:
: normally you call setActivationFunction() first and get a list of the used activation functions, which are set back with this function
Parameters:
actfunList the list of actfuns to be used

virtual void setSomeInternalParams ( bool  someInternalParams  )  [inline, virtual]

bool store ( FILE *  f  )  const [virtual]

stores the layer binary into file stream

Implements Storeable.

Reimplemented in Elman.

bool write ( FILE *  f  )  const

writes the layer ASCII into file stream (not in the storable interface)

Member Data Documentation

std::vector<matrix::Matrix> bias [protected]

matrix::Matrix bypassWeights [protected]

double eps

learning rate

bool initialised [protected]

matrix::Matrix input [protected]

std::vector<Layer> layers [protected]

std::vector<matrix::Matrix> smallids [protected]

bool someInternalParams [protected]

bool useBypass [protected]

std::vector<matrix::Matrix> weights [protected]

std::vector<matrix::Matrix> ys [protected]

std::vector<matrix::Matrix> zs [protected]

The documentation for this class was generated from the following files:
Generated on Fri Oct 30 16:29:02 2009 for Robot Simulator of the Robotics Group for Self-Organization of Control by  doxygen 1.4.7