darch (version 0.9.1)

backpropagation: Backpropagation learning function

Description

This function provides the backpropagtion algorithm for deep architectures.

Usage

backpropagation(darch,trainData,targetData,epoch)

Arguments

darch

An instance of the class DArch.

trainData

The data for training.

targetData

The targets for the data

epoch

Number of epochs for the training

Value

• The trained deep architecture

Details

The function is getting the learning parameters from the provided DArch object. It uses the attributes momentum, finalMomentum and momentumSwitch for the calculation of the new weights with momentum. The parameter epoch is provided for the change from momentum to finalMomentum and is compared to momentumSwitch. The attributes learnRateWeights and learnRateBiases will be used for updating the weights. To use the backpropagation function as the fine tuning function the layer functions of the darch DArch object must set to the versions which calculates also the derivatives of the function result.

References

Rumelhart, D., G. E. Hinton, R. J. Williams, Learning representations by backpropagating errors, Nature 323, S. 533-536, DOI: 10.1038/323533a0, 1986.

See Also

DArch rpropagation minimizeAutoencoder minimizeClassifier minimizeClassifier