#include <elman.h>
Inherits MultiLayerFFNN.
Inheritance diagram for Elman:
Public Member Functions | |
| Elman (double eps, const std::vector< Layer > &layers, bool useElman, bool useJordan=false, bool useBypass=false) | |
| virtual | ~Elman () |
| 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 (this will be different for every input, since it is a recurrent network) | |
| 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 | |
| virtual NetUpdate | weightIncrement (const matrix::Matrix &xsi) |
| determines the weight and bias updates | |
| virtual NetUpdate | weightIncrementBlocked (const matrix::Matrix &xsi_, int blockedlayer, int blockfrom, int blockto) |
| like weightIncrement but with blocked backprop flow for some neurons. | |
| virtual void | updateWeights (const NetUpdate &updates) |
| applies the weight increments to the weight (and bias) matrices with the learningrate and the learnRateFactor | |
| void | damp (double damping) |
| damps the weights and the biases by multiplying (1-damping) | |
| bool | store (FILE *f) const |
| stores the layer binary into file stream | |
| bool | restore (FILE *f) |
| restores the layer binary from file stream | |
| virtual paramkey | getName () const |
| return the name of the object | |
| 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. | |
Protected Attributes | |
| matrix::Matrix | elmanWeights |
| matrix::Matrix | elmanContext |
| matrix::Matrix | jordanWeights |
| matrix::Matrix | jordanContext |
| bool | useElman |
| bool | useJordan |
| Elman | ( | double | eps, | |
| const std::vector< Layer > & | layers, | |||
| bool | useElman, | |||
| bool | useJordan = false, |
|||
| bool | useBypass = false | |||
| ) | [inline] |
| eps | learning rate | |
| layers | Layer description (the input layer is not specified (always linear)) | |
| lambda | self-recurrent feedback strength of context neurons |
| virtual ~Elman | ( | ) | [inline, virtual] |
| void damp | ( | double | damping | ) | [virtual] |
| 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.
Reimplemented from MultiLayerFFNN.
| Inspectable::iparamvallist getInternalParams | ( | ) | const [virtual] |
| virtual paramkey getName | ( | ) | const [inline, virtual] |
| 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.
Reimplemented from MultiLayerFFNN.
| 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.
Reimplemented from MultiLayerFFNN.
| 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
Reimplemented from MultiLayerFFNN.
| const Matrix learn | ( | const matrix::Matrix & | input, | |
| const matrix::Matrix & | nom_output, | |||
| double | learnRateFactor = 1 | |||
| ) | [virtual] |
| const Matrix process | ( | const matrix::Matrix & | input | ) | [virtual] |
passive processing of the input (this will be different for every input, since it is a recurrent network)
Reimplemented from MultiLayerFFNN.
| bool restore | ( | FILE * | f | ) | [virtual] |
| bool store | ( | FILE * | f | ) | const [virtual] |
| void updateWeights | ( | const NetUpdate & | updates | ) | [virtual] |
applies the weight increments to the weight (and bias) matrices with the learningrate and the learnRateFactor
| NetUpdate weightIncrement | ( | const matrix::Matrix & | xsi | ) | [virtual] |
determines the weight and bias updates
| NetUpdate weightIncrementBlocked | ( | const matrix::Matrix & | xsi_, | |
| int | blockedlayer, | |||
| int | blockfrom, | |||
| int | blockto | |||
| ) | [virtual] |
like weightIncrement but with blocked backprop flow for some neurons.
| blockedlayer | index of layer with blocked neurons | |
| blockfrom | index of neuron in blockedlayer to start blocking | |
| blockto | index of neuron in blockedlayer to end blocking (if -1 then to end) (not included) |
matrix::Matrix elmanContext [protected] |
matrix::Matrix elmanWeights [protected] |
matrix::Matrix jordanContext [protected] |
matrix::Matrix jordanWeights [protected] |
bool useElman [protected] |
bool useJordan [protected] |
1.4.7