- initialize a feedforward neural net using the SCAWI
method
SYNOPSIS
int scawi_ffnet
(net, dset)
ARGUMENTS
NET
*net
The network to be initialized.
DATASET
*dset
The dataset to be used for initialization.
RETURNS
The function returns TRUE if an error was detected, if no error was detected
the function returns FALSE and the network is initialized.
FUNCTION
The Statistically Controlled Activation Weight Initialization (SCAWI) is a
neural network initialization method introduced by Drago and Ridella. This
method can be used to initialize a feedforward network in such a way that
it training performance is boosted.
NOTE
The dataset must be an IOSET (see DATASET-flags) and the network an
FFNET (see NET-flags). A detailed description of the SCAWI method
can be found in IEEE Transactions on Neural Networks, Vol. 3, No 4., July
1992.