The Complete SPRLIB & ANNLIB

PN_STRUCT

- a local structure for pseudo-Newton backpropagation learning (LOCAL)

SYNOPSIS

typedef struct pn_struct

{

 double Momentum Previous change in a weight; used in the momentum-term in the generalized delta-rule.

 double Accum_Delta_1 Sum of the delta's from (a part of) this neuron's delta-feeding units.

 double Accum_Delta_2 Sum of the "second-order delta's" from (a part of) this unit's delta-feeding units.

 int NrUpdates Number of times a certain weight is updated.

 int Changed Auxiliary for use with shared weights networks.

} PN_STRUCT;

DESCRIPTION

This structures is allocated for and attached to each weight and bias in the network by pn_init. Its fields are used during pseudo-Newton training. The connection is made through the WEIGHT structure's SystemData.p field (see system_data).

SEE ALSO

pn_learn

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