The Complete SPRLIB & ANNLIB

init_wavnet

-initialize a wavelet network

SYNOPSIS

int init_wavnet (net, dset, rand, L, U, comp)

ARGUMENTS

NET *net Wavelet network
DATASET *dset Dataset used for training the network
int rand Boolean indicating whether the weighs should be initialized randomly
double L Lower value for scaling the dataset prior to training
double U Upper value for scaling the dataset prior to training
int comp Parameter specifying the maximal amount of compression of a wavelon

FUNCTION

Initialization of the Wavenet is performed as originally proposed in Zhang/ Benveniste (1992). Here, the parameters are initialized as follows: 

  - g_0     := mean(outputs), where the observed outputs are used to 
               approximate the function f.
  - w_ji    := 0; 
  - t_i/s_i := some initialization procedure, described in Zhang/Benveniste. 
               First, the parameters are computed for the 1-dimensional 
               case (scalar inputs), next these parameters are compiled 
               into t_vectors and Di-matrices, consisting of the 
               proper "scalar values".

RETURNS

TRUE if an error was detected, FALSE otherwise.

NOTE

This function should be called before a wavelet network is being trained.

SEE ALSO

create_wavnet , learn_wavnet , free_wavnet

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