comparison toolboxes/FullBNT-1.0.7/netlab3.3/som.m @ 0:e9a9cd732c1e tip

first hg version after svn

+function net = som(nin, map_size)
+%SOM	Creates a Self-Organising Map.
+%
+%	Description
+%	NET = SOM(NIN, MAP_SIZE) creates a SOM NET with input dimension (i.e.
+%	data dimension) NIN and map dimensions MAP_SIZE.  Only two-
+%	dimensional maps are currently implemented.
+%
+%	The fields in NET are
+%	  type = 'som'
+%	  nin = number of inputs
+%	  map_dim = dimension of map (constrained to be 2)
+%	  map_size = grid size: number of nodes in each dimension
+%	  num_nodes = number of nodes: the product of values in map_size
+%	  map = map_dim+1 dimensional array containing nodes
+%	  inode_dist = map of inter-node distances using Manhatten metric
+%
+%	The map contains the node vectors arranged column-wise in the first
+%	dimension of the array.
+%
+%	See also
+%	KMEANS, SOMFWD, SOMTRAIN
+%
+
+%	Copyright (c) Ian T Nabney (1996-2001)
+
+net.type = 'som';
+net.nin = nin;
+
+% Create Map of nodes
+if round(map_size) ~= map_size | (map_size < 1)
+    error('SOM specification must contain positive integers');
+end
+
+net.map_dim = length(map_size);
+if net.map_dim ~= 2
+    error('SOM is a 2 dimensional map');
+end
+net.num_nodes = prod(map_size);
+% Centres are stored by column as first index of multi-dimensional array.
+% This makes extracting them later more easy.
+% Initialise with rand to create square grid
+net.map = rand([nin, map_size]);
+net.map_size = map_size;
+
+% Crude function to compute inter-node distances
+net.inode_dist = zeros([map_size, net.num_nodes]);
+for m = 1:net.num_nodes
+    node_loc = [1+fix((m-1)/map_size(2)), 1+rem((m-1),map_size(2))];
+    for k = 1:map_size(1)
+	for l = 1:map_size(2)
+	    net.inode_dist(k, l, m) = round(max(abs([k l] - node_loc)));
+	end
+    end
+end