new file: class2_pybrain.py

	new file:   class_pybrain.py
	modified:   iris_dataset.py
	modified:   neuralnets_functions.py

class2_pybrain.py

@@ -0,0 +1,30 @@
+import pybrain
+
+from pybrain.tools.shortcuts import buildNetwork
+from pybrain.datasets import SupervisedDataSet
+from pybrain.supervised import BackpropTrainer
+from pybrain.structure.modules import SigmoidLayer
+
+network = buildNetwork(2,3,1,outclass = SigmoidLayer, bias = False)
+
+dataset = SupervisedDataSet(2,1)
+dataset.addSample((0,0),(0,))
+dataset.addSample((0,1),(1,))
+dataset.addSample((1,0),(1,))
+dataset.addSample((1,1),(0,))
+
+optimizer = BackpropTrainer(module=network, dataset = dataset, learningrate = 0.3)
+epochs = 5000
+error = []
+for epoch in range(epochs):
+    error_average = optimizer.train()
+    if epoch % 1000 == 0:
+        print('Epoch: ' + str(epoch + 1) + ' Error: ' + str(error_average))
+        error.append(error_average)
+print('Epoch: ' + str(epoch + 1) + ' Error: ' + str(error_average))
+
+import matplotlib.pyplot as plt
+plt.xlabel('Epoch')
+plt.ylabel('Error')
+plt.plot(error)
+plt.show()

class_pybrain.py

@@ -0,0 +1,26 @@
+import pybrain
+
+from pybrain.structure import FeedForwardNetwork
+from pybrain.structure import SigmoidLayer, LinearLayer, BiasUnit
+from pybrain.structure import FullConnection
+
+network = FeedForwardNetwork()
+input_layer = LinearLayer(2)
+hidden_layer = SigmoidLayer(3)
+output_layer = SigmoidLayer(1)
+bias0 = BiasUnit()
+bias1 = BiasUnit()
+
+network.addModule(input_layer)
+network.addModule(hidden_layer)
+network.addModule(output_layer)
+network.addModule(bias0)
+network.addModule(bias1)
+
+input_to_hidden = FullConnection(input_layer, hidden_layer)
+hidden_to_output= FullConnection(hidden_layer, output_layer)
+bias_hidden = FullConnection(bias0, hidden_layer)
+bias_output = FullConnection(bias1, output_layer)
+
+network.sortModules()
+print(input_to_hidden.params)

iris_dataset.py

@@ -36,7 +36,6 @@
     input_x_delta0 = (input_layer.T).dot(delta_hidden_layer)
     weights0 = weights0 + (input_x_delta0*learning_rate)
 print('Epoch: ' + str(epoch + 1) + ' Error: ' + str(average))
-#print(inputs)
-#for i in range(len(inputs)):
-print(iris.target_names[int(round(nnf.sigmoid_output(50,weights0,weights1)))])
 
+for i in inputs:
+    print(iris.target_names[abs(int(round(nnf.sigmoid_output(i,weights0,weights1))))])

neuralnets_functions.py

@@ -19,6 +19,6 @@ def sigmoid_derivative(sigmoid):
     return sigmoid*(1 - sigmoid)
 
 def sigmoid_output(instance,weights0,weights1):
-  hidden_layer = sigmoid_derivative(np.dot(instance, weights0))
-  output_layer = sigmoid_derivative(np.dot(hidden_layer, weights1))
+  hidden_layer = sigmoid(np.dot(instance, weights0))
+  output_layer = sigmoid(np.dot(hidden_layer, weights1))
   return output_layer[0]