Learn R Programming
darch (version 0.10.0)

RBM-class: Class for restricted Boltzmann machines

Description

This class represents a restricted Boltzmann machine.

Arguments

Slots

learnRateBiasVisible

Object of class "numeric". Learning rate of the visible biases.

learnRateBiasHidden

Object of class "numeric". Learning rate of the hidden biases.

weightCost

Object of class "numeric". Weight cost for the update of the weights.

numHidden

Object of class "numeric". Number of hidden units.

numVisible

Object of class "numeric". Number of visible units.

weights

Object of class "matrix". Weight matrix.

weightInc

Object of class "matrix". Matrix of update values for the Weight.

output

Object of class "matrix". Output matrix of the RBM.

visibleBiases

Object of class "array". Visible biases array.

visibleBiasesInc

Object of class "array". Array of update values for the visible biases

visibleUnitFunction

Object of class "function". Unit function for the visible units.

visibleUnitStates

Object of class "list". States of the visible units.

hiddenBiases

Object of class "array". Hidden biases array.

hiddenBiasesInc

Object of class "array". Array of update values for the hidden biases.

hiddenUnitFunction

Object of class "function". Unit function for the hidden units.

hiddenUnitStates

Object of class "list". States of the hidden units.

updateFunction

Object of class "function". Function for updating the weights and biases.

posPhaseData

Object of class "list". Attribute to save the positive phase data during the training.

ffWeights

Object of class "ff_matrix". Weight ff matrix. Used when the ff attribute is TRUE.

ffOutput

Object of class "ff_matrix". Output ff matrix of the RBM. Used when the ff attribute is TRUE.

ffHiddenBiases

Object of class "ff_array". Hidden biases ff array. Used when the ff attribute is TRUE.

ffVisibleBiases

Object of class "ff_array". Hidden biases ff array. Used when the ff attribute is TRUE.

Details

For generating a RBM it is recommended to use the constructor function: newRBM. The RBM can be trained with the implementation of the contrastive divergence method trainRBM. The class inherits the attributes from the '>Net. The if the attribute ff is TRUE, the attributes with the ff-prefix are used to save the parameters of the RBM network. It is recommended to use the setter and getter method for access the attributes, because then there is no need to request the ff attribute to access the right attribute.

See Also

'>Net, '>DArch, trainRBM