Add a useful package (from Source forge) for octave

[CreaPhase.git] / octave_packages / nnet-0.1.13 / __init.m

diff --git a/octave_packages/nnet-0.1.13/__init.m b/octave_packages/nnet-0.1.13/__init.m
new file mode 100644 (file)
index 0000000..3105883
--- /dev/null
+++ b/octave_packages/nnet-0.1.13/__init.m
@@ -0,0 +1,84 @@
+## Copyright (C) 2005 Michel D. Schmid  <michaelschmid@users.sourceforge.net>
+##
+##
+## This program is free software; you can redistribute it and/or modify it
+## under the terms of the GNU General Public License as published by
+## the Free Software Foundation; either version 2, or (at your option)
+## any later version.
+##
+## This program is distributed in the hope that it will be useful, but
+## WITHOUT ANY WARRANTY; without even the implied warranty of
+## MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+## General Public License for more details.
+##
+## You should have received a copy of the GNU General Public License
+## along with this program; see the file COPYING.  If not, see
+## <http://www.gnu.org/licenses/>.
+
+## -*- texinfo -*-
+## @deftypefn {Function File} {} @var{net} = __init (@var{net})
+## @code{__init} initializes a neural network. This will be done
+## with the function @code{rand} from octave.
+##
+## @example
+## net = __init(net);
+## @end example
+##
+## This function takes the octave function "rand" to init the 
+## neural network weights.
+##
+## @noindent
+## @end deftypefn
+
+
+## Author: Michel D. Schmid
+
+function net=__init(net)
+
+  ## check number of inputs
+  error(nargchk(1,1,nargin));
+
+  ## check input
+  if ( !__checknetstruct(net) )
+    error("__init: wrong argument type, must be a structure!");
+  endif
+
+
+  if (strcmp(net.networkType,"newff"))
+
+    ## init with random numbers between +-1
+    ## input weight layer
+    mRand = rand(net.layers{1}.size,net.inputs{1}.size);
+    net.IW{1} = mRand*2-1;
+
+    ## hidden layers
+    nLayers = net.numLayers;
+    for i=2:nLayers
+      mRand = rand(net.layers{i}.size,net.layers{i-1}.size);
+      net.LW{i,i-1} = mRand*2-1;
+    endfor
+    for i=1:nLayers
+      mRand = rand(net.biases{i}.size,1);
+      net.b{i} = mRand*2-1;
+    endfor
+  elseif (strcmp(net.networkType,"newp"))
+
+    ## init with zeros
+    inputRows = size(net.inputs{1,1}.range,1);
+    net.IW{1} = zeros(inputRows,1);
+    net.b{1} = zeros(1,1);
+  endif
+
+  ## warn user of constant inputs
+  for i=1:net.numInputs
+    prange = net.inputs{i}.range;
+    if (any(prange(:,1) == prange(:,2)))
+      fprintf("\n")
+      fprintf("** Warning in INIT\n")
+      fprintf("** Network net.inputs{%g}.range has a row with equal min and max values.\n",i)
+      fprintf("** Constant inputs do not provide useful information.\n")
+      fprintf("\n")
+    end
+  end
+
+endfunction