grossberg2.conf

% true oznaczałoby losowanie wag dla neuronów, weights jest wtedy ignorowane
random_weights = true;

% dane wejściowe
input_data = [
 0 0 0;
 0 0 1;
 0 1 0;
 0 1 1;
 1 0 0;
 1 0 1;
 1 1 0;
 1 1 1;
];

expected = [
 0;
 1;
 1;
 0;
 1;
 0;
 0;
 1;
];

%global epochs = [1 8000 16000 24000]

global layers = {};
learn_steps = 10000;
% pierwsza warstwa
% liczba neuronów w warstwie
layers{1}.type='kohonen';
layers{1}.neurons = 8;
layers{1}.row_count = 1;
% funkcja aktywacji
layers{1}.activation_function =@(X) sigmoid(X);
% wartość bias
layers{1}.bias = 0;
% przedział, z którego są losowane wagi
layers{1}.rand_min = -1.0;
layers{1}.rand_max = 1.0;
layers{1}.neighbourhood_width = [0 0 0 0 0];%[7.0 5.0 3.0 1.0];
layers{1}.conscience_coefficient = [1.0 0.5 0.25 0.125 0]; %[0 0 0 0];%
layers{1}.learning_coefficient = [0.06 0.03 0.015 0.0075 0];
layers{1}.learn = false;
layers{1}.random_weights = false;

layers{1}.weights = [
   8.7557e-02   4.7263e-04  -5.2823e-10   2.2734e-11;
   8.2916e-01   5.7302e-01   5.7544e-01   5.7304e-01;
  -8.3701e-01   1.0731e-01   6.4703e-01   5.9212e-01;
  -5.3046e-01   2.4351e-04   9.9924e-01   1.6416e-03;
  -2.2320e-01   7.0637e-01   2.2126e-03   7.0668e-01;
   1.1293e-01   9.9954e-01  -3.3158e-07  -5.7041e-07;
  -9.7452e-02   3.4252e-01   5.1528e-01   3.4149e-01;
  -6.2274e-01   1.8204e-02   7.0302e-01   7.0302e-01;
  ];
layers{1}.epochs = [1 1000 2000 2500 3000];


layers{2}.type='grossberg';
layers{2}.neurons=1;
layers{2}.bias = 0;
layers{2}.rand_min = -1.0;
layers{2}.rand_max = 1.0;
layers{2}.activation_function = @(X) linear(X,1,0);
layers{2}.epochs = [3000 10000 15000 20000];
layers{2}.coeffs = [0.8 0.5 0.3 0.1];
layers{2}.epoch=0;
layers{2}.random_weights = false;
layers{2}.rule = 'delta'; %'widrow';
layers{2}.learn = false;
layers{2}.weights = [
-3.1242e-02  -1.4822e-323    1.0000e+00    3.9041e-01    1.0000e+00   1.4822e-323    1.0000e+00    4.3799e-01   1.4822e-323
];