1 files changed, 86 insertions, 0 deletions
diff --git a/2017/ev3-controller/nn_11_function.py b/2017/ev3-controller/nn_11_function.py
new file mode 100644
index 0000000..677db1b
--- /dev/null
+++ b/2017/ev3-controller/nn_11_function.py
@@ -0,0 +1,86 @@
+import numpy as np
+from data_processing import process
+
+#get the data
+training_x, training_y, test_x, test_y = process()
+X=np.array(training_x)
+y=np.array(training_y)
+
+
+#sigmoid function
+def nonlin(x, deriv=False):
+    if (deriv == True):
+        return x * (1 - x)
+    return 1 / (1 + np.exp(-x))
+
+#print(X)
+#print(y)
+
+def train(X=X,y=y, descent_rate=0.01, hl1=2):
+    np.random.seed(1)
+
+    # randomly initialize our weights with mean 0
+    syn0 = 2 * np.random.random((len(X[1]), hl1)) - 1
+    syn1 = 2 * np.random.random((hl1, len(y[1]))) - 1
+
+    for j in range(5000):
+
+        # Feed forward through layers 0, 1, and 2
+        l0 = X
+        l1 = nonlin(np.dot(l0, syn0))
+        l2 = nonlin(np.dot(l1, syn1))
+
+        #print(l2)
+
+        # calculate the error in from the target value
+        l2_error = y - l2
+
+        if (j % 100) == 0:
+            print  ( "Error:" + str(np.mean(np.abs(l2_error))))
+
+        # calculate the desired change in weights - bigger if our confidence is higher, use descent rate to control convergence
+        l2_delta = l2_error * nonlin(l2, deriv=True)*descent_rate
+
+        # how much did each l1 value contribute to the l2 error (according to the weights)?
+        l1_error = l2_delta.dot(syn1.T)
+
+        # calculate the desired change in weights - bigger if our confidence is higher, use descent rate to control convergence.
+        l1_delta = l1_error * nonlin(l1, deriv=True)*descent_rate
+
+        #update the weights
+        syn1 += l1.T.dot(l2_delta)
+        syn0 += l0.T.dot(l1_delta)
+
+
+    return syn0, syn1
+
+
+def calculate_output(syn0, syn1, l0):
+    l1 = nonlin(np.dot(l0, syn0))
+    l2 = nonlin(np.dot(l1, syn1))
+    return l2
+
+
+#test neural network
+l0=test_x
+#print(l0)
+
+#train the neural network
+syn0, syn1=train(X, y)
+print(syn0)
+print(syn1)
+
+#get the output values based on the trained network
+l2 = calculate_output(syn0, syn1, l0)
+
+#check correctness of nn
+correct = np.equal(np.argmax(test_y, 1), np.argmax(l2, 1))
+print(correct)
+
+accuracy = np.mean(correct.astype(float))
+print(accuracy)
+
+#Print inputs and predicted outputs
+#for i in range(len(l0)):
+    #print(l0[i])
+    #print(l2[i])
+\ No newline at end of file