function [w1,w2] = initlvq(p,s1,t)
%INITLVQ Inititialize LVQ network.
%  
%  This function is obselete.
%  Use NNT2LVQ and INIT to update and initialize your network.

nntobsf('initlvq','Use NNT2LVQ and INIT to update and initialize your network.')

%  [W1,W2] = INITLVQ(P,S1,T)
%    P  - RxQ matrix of input vectors.
%    S1 - Size of the competitive hidden layer.
%    T  - S2xQ matrix of target vectors.
%  Returns:
%    W1 - S1xR weight matrix for competitive layer.
%    W2 - S2xS1 weight matrix for linear layer.
%  
%  IMPORTANT: Each ith row of P must contain expected
%    min and max values for the ith input.
%  
%  EXAMPLE: P = [-1 2 3 2 -2 1 0; 2 -1 1 -3 1 -2 2];
%           T = [0 0 0 1 1 1 1; 1 1 1 0 0 0 0];
%           [W1,W2] = initlvq(P,7,T)
%           p = [1; -1];
%           a = simulvq(p,W1,W2)
%  
%  See also NNINIT, LVQ, SIMULVQ, TRAINLVQ.

% Mark Beale, 12-15-93
% Copyright 1992-2002 The MathWorks, Inc.
% $Revision: 1.11 $  $Date: 2002/04/14 21:14:08 $

if nargin < 3,error('Not enough input arguments'),end

[r,q] = size(p);
if (q < 2), error('First argument has less than 2 columns.'),end
[s2,q] = size(t);

% CLASSES -> TARGETS
if (s2 == 1)
  t = ind2vec(t);
  [s2,q] = size(t);
end

% LAYER 1
w1 = midpoint(s1,p);

% LAYER 2
perc = nnsumr(t)/q;
ind = round(cumsum(perc)*s1);

start_ind = 1;
class_ind = zeros(1,s1);
for i=1:s2
  stop_ind = ind(i);
  class_ind(start_ind:stop_ind) = i*ones(1,stop_ind-start_ind+1);
  start_ind = stop_ind+1;
end
w2 = ind2vec(class_ind);