TrainingAlgorithm

TrainingAlgorithm

TrainingAlgorithm is a property of a NeuralNet object. It specifies the training algorithm used when training the net. It can be one of BackpropagationIncremental, BackpropagationBatch, Rprop or Quickprop. See each algorithm for further information. The default algorithm is Rprop.

BackpropagationBatch

BackpropagationBatch is a value of the NeuralNet property TrainingAlgorithm. The algorithm updates the weights in the net after training on all input/output pairs using standard back propagation.

BackpropagationIncremental

BackpropagationIncremental is a value of the NeuralNet property TrainingAlgorithm. The algorithm updates the weights in the net after training on each input/output pair using standard back propagation.

Quickprop

Quickprop is a value of the NeuralNet property TrainingAlgorithm. The algorithm updates the weights in the net after training on all input/output pairs using an advanced batch algorithm. It uses the learning rate along with the Quickprop... properties.

Rprop

Rprop is a value of the NeuralNet property TrainingAlgorithm. The algorithm updates the weights in the net after training on all input/output pairs using an adaptive algorithm. It does not use the learning rate. It uses the RpropStepsize... properties. It is the default algorithm.

Created by freegoldbar  (September 16, 2004)