Created main and configuration file

configuration.m

@@ -0,0 +1,13 @@
+15;
+data_file = 'terrain/terrain03.data';
+starting_line = 2;
+beta = 15;
+activation = 'exp';
+net = [2, 45, 50, 1];
+eta = 0.1;
+max_iterations = 1e2;
+cut_error = 1e-4;
+alpha = 0;
+adaptative_eta = true;
+print_error = true;
+print_estimation = false;

main.m

@@ -0,0 +1,37 @@
+1;
+source multilayer_perceptron.m
+source configuration.m
+
+terrain = dlmread(data_file);
+terrain = terrain(starting_line:end,:);
+
+if activation == 'exp'
+  [f,fder] = activation_exp(beta);
+else 
+  [f,fder] = activation_tanh(beta);
+end
+
+[weights, output, mse] = multilayer_perceptron_learn(terrain(:,1:2)', terrain(:,3)', net, f, fder, eta, max_iterations, cut_error, alpha, adaptative_eta);
+
+if print_error
+  figure(1);
+  semilogy(mse, 'LineWidth', 2);
+  title("ECM");
+  xlabel("Épocas");
+  ylabel("ECM en escala logarítmica");
+end
+
+if print_estimation
+  figure(2);
+  x = [-3:0.025:3];
+  y = x;
+  for i = 1:length(x)
+          for j = 1:length(y)
+              z(i,j) = get_output([x(i);y(j)],weights,net,f); 
+          end
+  end
+  surf(x, y, z);
+  mymap = [0, 0, 0.6; 0.2, 0.8, 0; 0.4, 0.8, 0; 0.5, 0.8, 0; 0.7, 0.7, 0; 0.8, 0.8, 0.8; 1, 1, 1];
+  colormap(get_map(mymap));
+  title("Aproximación de la función");
+end

multilayer_perceptron.m

@@ -21,10 +21,9 @@
                                     learning_factor=.5, max_iterations=1000, tolerance=1e-5, alpha=0, adaptative_eta=false, dbug=false)
     %number of entries
     n = length(entries(1,:));
-
+    eta = learning_factor;
     %number of layers
     m = 0;
-    a = alpha;
     %setup
     for i = 2:length(neurons_per_layer)
         m = m + 1;
@@ -49,7 +48,7 @@
             if dbug 
                 layer_entry
                 fflush(1);
-	        end
+	          end
             h{M} = weights{M} * layer_entry{M}';
 
             %no linear
@@ -74,8 +73,8 @@
                 prev_delta_w = delta_w;
             end
             for i = 1:M
-              if iteration > 1 
-                delta_w{i} = learning_factor * d{i}' * layer_entry{i} + a * delta_w{i};
+              if iteration > 1 && a ~= 0
+                delta_w{i} = -learning_factor * d{i}' * layer_entry{i} + a * delta_w{i};
               else 
                 delta_w{i} = learning_factor * d{i}' * layer_entry{i};
               end
@@ -85,29 +84,29 @@
         %get iteration error
         error_per_iteration(iteration) = sum((expected_output - output).^2)/n;
         if adaptative_eta
-          if iteration > 10 
-            if error_per_iteration(iteration) > error_per_iteration(iteration - 10)
+          if iteration > 1 
+            if error_per_iteration(iteration) > error_per_iteration(iteration - 1)
               weights = old_weigths;
+              error_per_iteration(iteration) = error_per_iteration(iteration - 1);
               delta_w = prev_delta_w;
               learning_factor = 0.9 * learning_factor;
-              a = 0;
+              a = 0;    
             else 
               a = alpha;
-              learning_factor = 1.1 * learning_factor;
+              learning_factor = 0.1 * eta + learning_factor;
             end
+          else
+            a = alpha;
           end
         end
-        [error_per_iteration(iteration),iteration,toc]
+        [error_per_iteration(iteration),iteration,toc, learning_factor]
         fflush(1);
         if error_per_iteration(iteration) <= tolerance
             return
         end
-        %if error_per_iteration(iteration) <= 5e-4
-         %   learning_factor = .02;
-        %end
-        %if error_per_iteration(iteration) <= 4e-4
-         %   learning_factor = .01;
-        %end
+        if(learning_factor < 0.001 * eta)
+          learning_factor = 0.1 * eta;
+        end
     end
 
 end

plotter.m

@@ -1,4 +1,6 @@
 source multilayer_perceptron.m
+x = [-5:0.01:5];
+y = x;
 for i = 1:length(x)
         for j = 1:length(y)
             z(i,j) = get_output([x(i);y(j)],weights,net,f); 
@@ -7,5 +9,7 @@
 
 plot3(terrain(:,2),terrain(:,1),terrain(:,3),'.','markersize',12)
 hold on;
+mymap = [0,0,0.6;0.2,0.8,0;0.4,0.8,0;0.5,0.8,0;0.7,0.7,0;0.8,0.8,0.8;1,1,1];
+colormap(get_map(mymap));
 surf(x,y,z);
 hold off;

test.m

@@ -2,9 +2,9 @@
 source multilayer_perceptron.m
 terrain = dlmread('terrain/terrain03.data');
 terrain = terrain(2:end,:);
-[f,fder] = activation_exp(1);
+[f,fder] = activation_exp(15);
 net = [2,45,50,1];
-[weights,output,mse] = multilayer_perceptron_learn(terrain(:,1:2)',terrain(:,3)',net,f,fder,.01,1e3,1e-4,0.9,true);
+[weights,output,mse] = multilayer_perceptron_learn(terrain(:,1:2)',terrain(:,3)',net,f,fder,.1,1e4,1e-4,0,true);
 x = [-3:0.01:3];
 y = x;
 %dlmwrite('weights.csv',cell2mat(weights));
@@ -15,6 +15,7 @@
         end
 end
 
+
 plot3(terrain(:,2),terrain(:,1),terrain(:,3),'.','markersize',12)
 hold on;
 mymap = [0,0,0.6;0.2,0.8,0;0.4,0.8,0;0.5,0.8,0;0.7,0.7,0;0.8,0.8,0.8;1,1,1];