
MATLAB compatible MEX and .m code for implementing FIR neural
networks.  (9/95)

Eric A. Wan 
Alex T. Nelson
Thyagarajan Srinivasan  

Oregon Graduate Institute

Use of software is free for academic institutions.  Businesses who find
this code useful should send a check for $200 payable to "Eric Wan -
OGI".

Eric Wan
Oregon Graduate Institute
P.O. Box 91000
Portland, OR 97291

ericwan@eeap.ogi.edu


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

CODE (Rev. 1.0):

weight_init.m    : Initializes weights for a standard or FIR network

ffnet2.c         : Simulates a standard 2 layer feedforward network with
                   optional taps at input. (tanh-linear)

firnet2.c        : Simulates a 2 layer FIR network (tanh-linear)

firnet3.c        : Simulates a 3 layer FIR network (tanh-tanh-linear)

bp_ffnet2.c      : Stochastic or batch gradient training of ffnet2

bp_firnet2.c     : Stochastic or batch gradient training of firnet2

bp_firnet3.c     : Stochastic or batch gradient training of firnet3

iterate_ffnet2.m : Iterated predictions for ffnet2

demo1ff2.m       : Demo 1 for ffnet2


*.mexsol         : Compiled files for SunOS 5.3.  To recompile on a different
                   system enter:  cmex filename.c.  MATLAB necessary.







