The train command will cause a training object to commence training the target ANN with the specified parameters.
BPTrainer, GATrainer, KohonenTrainer, LWFTrainer
Items marked with a * are optional.
for BPTrainer objects:
train alias TrainerAlias data TrainingData epochs #Epochs LearningRate* LearningRate Momentum* Momentum terminating* TerminatingError mode* TrainingMode
for LWFTrainer objects or to train an MLP with LWF:
train alias TrainerAlias data TrainingData epochs #Epochs LearningRate* LearningRate Momentum* Momentum decay* DecayRate terminating* TerminatingError mode* TrainingMode
for GATrainer objects:
train alias TrainerAlias data TrainingData generations #Generations crossover* #CrossoverPoints mutation* MutationRate selection* SelectionStrategy elitism* #Elites terminating* TerminatingError normalisation* FitnessNormalisation
for KohonenTrainer objects:
train alias TrainerAlias data TrainingData LearningRate LearningRate neighbourhood { Neighbourhood }
where:
alias is the alias of the trainer object
data is the alias of the training data object
epochs is the number of backpropagation epochs to train for
LearningRate is the backpropagation learning rate parameter
Momentum is the backpropagation momentum parameter
terminating specifies the terminating RMS error
mode is the training mode, either batch or patterm
decay is the weight decay rate parameter for LWF training
generations is the number of generations for GA training
crossover is the number of crossover points to use
mutation is the mutation rate to use in reproduction
selection is the selection strategy to use, either Roulette or Tournament
elitism specifies the number of elite individuals to preserve from
one generation to the next
normalisation specifies whether or not to use fitness normalisation,
either on or off
neighbourhood is an array of the activation neighourhood for Kohonen
training. The number of entries must be the same as the dimensionality
of the output map
This page is maintained by Michael Watts (http://mike.watts.net.nz)
Last modified on: 12/10/98.