finish 2_BasicModels

examples/2_BasicModels/rand.py

@@ -0,0 +1,14 @@
+import numpy as np
+rng = np.random
+print rng.randn()
+print rng.randn(1)
+print rng.randn(5)
+print rng.randn(1, 1)
+print rng.randn(2, 3)
+
+import tensorflow as tf
+# illegal print tf.truncated_normal()
+# illegal print tf.truncated_normal(1)
+print tf.truncated_normal([ 1 ])
+print tf.truncated_normal([ 2, 2 ])
+

examples/2_BasicModels/test.py

@@ -0,0 +1,38 @@
+import tensorflow as tf
+
+a = tf.placeholder(tf.int32, [None, 4])
+b = tf.placeholder(tf.int32, [4])
+
+c = a - b
+
+feed = {
+	a: [ [ 2, 2, 2, 2 ], [ 3, 3, 3, 3 ], [ 4, 4, 4, 4 ] ],
+	b: [ 1, 1, 1, 1 ]
+}
+sess = tf.Session()
+print(sess.run(c, feed_dict = feed))
+
+print 'reduce to 1'
+e = tf.reduce_sum(c, reduction_indices = 1)
+print(sess.run(e, feed_dict = feed))
+
+print 'reduce to 0'
+d = tf.reduce_sum(c, reduction_indices = 0)
+print(sess.run(d, feed_dict = feed))
+
+
+print 'reduce to [0, 1]'
+d = tf.reduce_sum(c, reduction_indices = [0, 1])
+print(sess.run(d, feed_dict = feed))
+
+print 'reduce to [1, 0]'
+d = tf.reduce_sum(c, reduction_indices = [1, 0])
+print(sess.run(d, feed_dict = feed))
+
+print 'reduce to [0, 0]'
+d = tf.reduce_sum(c, reduction_indices = [0, 0])
+print(sess.run(d, feed_dict = feed))
+
+print 'reduce to [1, 1]'
+d = tf.reduce_sum(c, reduction_indices = [1, 1])
+print(sess.run(d, feed_dict = feed))

examples/2_BasicModels/zuilinjin.py

@@ -0,0 +1,40 @@
+from __future__ import print_function
+
+import tensorflow as tf
+import numpy as np
+from tensorflow.examples.tutorials.mnist import input_data
+
+mnist = input_data.read_data_sets('MNIST_data', one_hot = True)
+
+train_image_set, train_label_set = mnist.train.next_batch(500)
+test_image_set, test_label_set = mnist.test.next_batch(10)
+
+
+# placeholder, Graph Input
+train_images = tf.placeholder(tf.float32, [ None, 784 ])
+test_image = tf.placeholder(tf.float32, [ 784 ])
+
+
+# L1 distance
+# the - will adapts different dimensions and than will iterates the train_images to minus test_image one by one
+distance = tf.reduce_sum(tf.abs(train_images - test_image), reduction_indices = 1)
+
+# find the nearest training example, Graph, not np.argmin
+predict_index = tf.arg_min(distance, 0)
+
+init = tf.initialize_all_variables()
+
+sess = tf.Session()
+sess.run(init)
+
+accuracy = 0
+for i in range(len(test_image_set)):
+	best_neighbour_index = sess.run(predict_index, feed_dict={ train_images: train_image_set, test_image: test_image_set[i] })
+
+	#print("test", i, "Prediction:", train_label_set[best_neighbour_index], "True Class:", test_label_set[i])
+	# vector representation --> index e.g. [ 0, 1, 0, 0] --> 1, [ 0, 0, 0, 1 ] --> 3	
+	print("test", i, "Prediction:", np.argmax(train_label_set[best_neighbour_index]), "True Class:", np.argmax(test_label_set[i]))
+	if np.argmax(train_label_set[best_neighbour_index]) == np.argmax(test_label_set[i]):
+		accuracy += 1. / len(test_image_set)
+
+print("Done!", "Accuracy:", accuracy)