The Complete SPRLIB & ANNLIB

samann_init

- initialise and set up a network for training with the SAMANN backpropagation learning rule

SYNOPSIS

int samann_init (net, set, scale)

ARGUMENTS

NET *net A pointer to a multilayer feedforward NET.
DATASET *set A pointer to the DATASET to be used to train the network.
int scale A boolean indicating whether the dataset set should be scaled or not.

RETURNS

TRUE if an error was detected, FALSE if no error was detected.

FUNCTION

After some initial checking, space is allocated for storing information in the SystemData field of the weights and biases (thetas). The pointer p in these structures point to SAMANN_STRUCT structures, used during the training process. If scale is TRUE, the dataset is scaled in such a way that the maximum distance between two samples is equal to the maximum distance in the map (i.e., sqrt(dim)). See also samann_scale_dataset. This is really only necessary when the network has sigmoid output units. Finally, the Sammon constant lambda is calculated (see samann_lambda).

NOTE

The samann_free function must be used after training to free the extra structures.

SEE ALSO

samann_scale_dataset, samann_lambda, samann_free, samann_learn

This document was generated using api2html on Thu Mar 5 09:00:00 MET DST 1998