Merge pull request #1 from Naveendurai/master

Init commit 🔆

Author: swaathi

01_basics.py

@@ -0,0 +1,36 @@
+# Basics of Neural Networking
+import numpy as np
+
+no_inputs = 5
+x_data = np.array([1, 2, 3, 4, 5])
+y_data = np.array([2, 4, 6, 8, 10])
+weight = 1
+bias = 0.0
+learning_rate = 0.01
+
+# The basic formula is Y = ( w * X ) + B
+# Y is the outputs
+# w is the weight
+# X is the inputs
+# B is the bias
+
+for i in range(10):
+    print 'Step', i, ':'
+    output = x_data * weight + bias
+    # [ 0.3  0.6  0.9  1.2  1.5]
+
+    difference = y_data - output
+
+    gradient = 0
+    for i in range(no_inputs):
+        gradient += difference[i] * x_data[i]
+
+    gradient /= no_inputs
+
+    loss = (difference * difference) / no_inputs
+    print 'Loss:', np.sum(loss)
+
+    weight = weight + (learning_rate * gradient)
+
+    print 'Weight: ', weight
+    print ''

02_tf_basics.py

@@ -0,0 +1,10 @@
+# Basics of Tensorflow
+
+import tensorflow as tf
+
+x_input = tf.constant([1, 2, 3, 4])
+# print x_input
+# Tensor("Const:0", shape=(4,), dtype=int32)
+
+sess = tf.Session()
+print sess.run(x_input)

03_tf_network.py

@@ -0,0 +1,27 @@
+# Basics NeuralNetwork of Tensorflow
+import tensorflow as tf
+import numpy as np
+x_data = tf.constant([1.0, 2.0, 3.0, 4.0, 5.0])
+y_data = tf.constant([2.0, 4.0, 6.0, 8.0, 10.0])
+
+x = tf.placeholder(tf.float32)
+y = tf.placeholder(tf.float32)
+
+weight = tf.Variable([1.0])
+bias = tf.Variable([0.0])
+
+init = tf.initialize_all_variables()
+predict = (x * weight) 
+loss = tf.reduce_mean(tf.square(predict - y) ) 
+optimizer = tf.train.GradientDescentOptimizer(0.01).minimize(loss/2)
+steps = 300
+with tf.Session() as sess:
+    sess.run(init)
+    X = sess.run(x_data)
+    Y = sess.run(y_data)
+    for i in range(steps):
+    	print 'Step: ', i, ':'
+    	print 'Loss: ', sess.run(loss, {x: X, y: Y})
+    	sess.run(optimizer, {x: X, y: Y})
+    print 'THE FINAL WEIGHT IS', sess.run(weight)[0]
+

04_tf_with_hidden.py

@@ -0,0 +1,39 @@
+# Tensorflow Basics on NeuralNetwork with hidden layer
+
+import tensorflow as tf
+import numpy as np
+
+INPUT_NEURONS = 2
+HIDDEN_LAYER_1_NEURONS = 4
+OUTPUT_NEURONS = 1
+x_data = np.array([[0, 0], [0, 1], [1, 1]])
+
+y_data = np.array([0, 1, 1])
+
+x = tf.placeholder(tf.float32)
+y = tf.placeholder(tf.float32)
+
+h1_layer = {'weight': tf.Variable(np.random.rand(INPUT_NEURONS, HIDDEN_LAYER_1_NEURONS).astype(np.float32)),
+            'bias': tf.constant(0.0)}
+output_layer = {'weight': tf.Variable(np.random.rand(HIDDEN_LAYER_1_NEURONS, OUTPUT_NEURONS).astype(np.float32)),
+                'bias': tf.constant(0.0)}
+
+
+h1 = tf.matmul(x, h1_layer['weight']) + h1_layer['bias']
+h1 = tf.sigmoid(h1)
+
+predict = tf.matmul(h1, output_layer['weight']) + output_layer['bias']
+
+loss = tf.reduce_mean(tf.square(predict - y))
+optimizer = tf.train.GradientDescentOptimizer(0.00001).minimize(loss)
+
+sess = tf.Session()
+init = tf.initialize_all_variables()
+sess.run(init)
+
+for i in range(50000):
+		if i % 5000 == 0:
+			print 'Loss: ', sess.run(loss, {x: x_data, y: y_data})
+		sess.run(optimizer, {x: x_data, y: y_data})
+
+print sess.run(predict, {x: [[1, 1]] })