Skip to content

Commit 39f541d

Browse files
wiibrewwiibrew
committed
adding notebook
1 parent 8d79a11 commit 39f541d

File tree

4 files changed

+573
-0
lines changed

4 files changed

+573
-0
lines changed

‎course_2_tf_nn.ipynb‎

Lines changed: 190 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,190 @@
1+
{
2+
"cells": [
3+
{
4+
"cell_type": "code",
5+
"execution_count": 2,
6+
"metadata": {
7+
"collapsed": false
8+
},
9+
"outputs": [
10+
{
11+
"name": "stdout",
12+
"output_type": "stream",
13+
"text": [
14+
"Extracting ./mnist/train-images-idx3-ubyte.gz\n",
15+
"Extracting ./mnist/train-labels-idx1-ubyte.gz\n",
16+
"Extracting ./mnist/t10k-images-idx3-ubyte.gz\n",
17+
"Extracting ./mnist/t10k-labels-idx1-ubyte.gz\n",
18+
"('Epoch:', '0001', 'cost=', '215.548141965')\n",
19+
"('Epoch:', '0002', 'cost=', '54.977557694')\n",
20+
"('Epoch:', '0003', 'cost=', '33.899888993')\n",
21+
"('Epoch:', '0004', 'cost=', '23.234023376')\n",
22+
"('Epoch:', '0005', 'cost=', '16.552313167')\n",
23+
"('Epoch:', '0006', 'cost=', '12.184614655')\n",
24+
"('Epoch:', '0007', 'cost=', '8.918999288')\n",
25+
"('Epoch:', '0008', 'cost=', '6.555203167')\n",
26+
"('Epoch:', '0009', 'cost=', '4.864825427')\n",
27+
"('Epoch:', '0010', 'cost=', '3.541727996')\n",
28+
"('Epoch:', '0011', 'cost=', '2.601980731')\n",
29+
"('Epoch:', '0012', 'cost=', '2.013708151')\n",
30+
"('Epoch:', '0013', 'cost=', '1.447752024')\n",
31+
"('Epoch:', '0014', 'cost=', '1.284220558')\n",
32+
"('Epoch:', '0015', 'cost=', '1.063494972')\n",
33+
"('Epoch:', '0016', 'cost=', '1.089214503')\n",
34+
"('Epoch:', '0017', 'cost=', '0.819465103')\n",
35+
"('Epoch:', '0018', 'cost=', '0.826465986')\n",
36+
"('Epoch:', '0019', 'cost=', '0.756363073')\n",
37+
"('Epoch:', '0020', 'cost=', '0.756904836')\n",
38+
"('Epoch:', '0021', 'cost=', '0.772401051')\n",
39+
"('Epoch:', '0022', 'cost=', '0.591537078')\n",
40+
"('Epoch:', '0023', 'cost=', '0.518754110')\n",
41+
"('Epoch:', '0024', 'cost=', '0.653424654')\n",
42+
"('Epoch:', '0025', 'cost=', '0.639180361')\n",
43+
"('Epoch:', '0026', 'cost=', '0.418257485')\n",
44+
"('Epoch:', '0027', 'cost=', '0.434976982')\n",
45+
"('Epoch:', '0028', 'cost=', '0.606400410')\n",
46+
"('Epoch:', '0029', 'cost=', '0.475488307')\n",
47+
"('Epoch:', '0030', 'cost=', '0.458589170')\n",
48+
"Optimization Finished!\n",
49+
"('Accuracy:', 0.96039999)\n"
50+
]
51+
}
52+
],
53+
"source": [
54+
"#get the mnist data \n",
55+
"# wget http://deeplearning.net/data/mnist/mnist.pkl.gz\n",
56+
"\n",
57+
"\n",
58+
"\n",
59+
"\n",
60+
"from tensorflow.examples.tutorials.mnist import input_data\n",
61+
"mnist = input_data.read_data_sets(\"./mnist/\", one_hot=True)\n",
62+
"\n",
63+
"import tensorflow as tf\n",
64+
"\n",
65+
"# Parameters\n",
66+
"learning_rate = 0.001\n",
67+
"training_epochs = 30\n",
68+
"batch_size = 100\n",
69+
"display_step = 1\n",
70+
"\n",
71+
"# Network Parameters\n",
72+
"n_hidden_1 = 256 # 1st layer number of features\n",
73+
"n_hidden_2 = 512 # 2nd layer number of features\n",
74+
"n_input = 784 # MNIST data input (img shape: 28*28)\n",
75+
"n_classes = 10 # MNIST total classes (0-9 digits)\n",
76+
"\n",
77+
"# tf Graph input\n",
78+
"x = tf.placeholder(\"float\", [None, n_input])\n",
79+
"y = tf.placeholder(\"float\", [None, n_classes])\n",
80+
"\n",
81+
"\n",
82+
"# Create model\n",
83+
"def multilayer_perceptron(x, weights, biases):\n",
84+
" # Hidden layer with RELU activation\n",
85+
" layer_1 = tf.add(tf.matmul(x, weights['h1']), biases['b1'])\n",
86+
" layer_1 = tf.nn.relu(layer_1)\n",
87+
" # Hidden layer with RELU activation\n",
88+
" layer_2 = tf.add(tf.matmul(layer_1, weights['h2']), biases['b2'])\n",
89+
" layer_2 = tf.nn.relu(layer_2)\n",
90+
"\n",
91+
" # layer_3 = tf.add(tf.matmul(layer_2, weights['h3']), biases['b3'])\n",
92+
" # layer_3 = tf.nn.relu(layer_3)\n",
93+
"\n",
94+
"\n",
95+
"\n",
96+
" #we can add dropout layer\n",
97+
" # drop_out = tf.nn.dropout(layer_2, 0.75)\n",
98+
"\n",
99+
"\n",
100+
"\n",
101+
" # Output layer with linear activation\n",
102+
" out_layer = tf.matmul(layer_2, weights['out']) + biases['out']\n",
103+
" return out_layer\n",
104+
"\n",
105+
"# Store layers weight & biases\n",
106+
"weights = {\n",
107+
" #you can change \n",
108+
" 'h1': tf.Variable(tf.random_normal([n_input, n_hidden_1])),\n",
109+
" 'h2': tf.Variable(tf.random_normal([n_hidden_1, n_hidden_2])),\n",
110+
" #'h3': tf.Variable(tf.random_normal([n_hidden_1, n_hidden_2])),\n",
111+
" 'out': tf.Variable(tf.random_normal([n_hidden_2, n_classes]))\n",
112+
"}\n",
113+
"biases = {\n",
114+
" 'b1': tf.Variable(tf.random_normal([n_hidden_1])),\n",
115+
" 'b2': tf.Variable(tf.random_normal([n_hidden_2])),\n",
116+
" #'b3': tf.Variable(tf.random_normal([n_hidden_2])),\n",
117+
" 'out': tf.Variable(tf.random_normal([n_classes]))\n",
118+
"}\n",
119+
"\n",
120+
"# Construct model\n",
121+
"pred = multilayer_perceptron(x, weights, biases)\n",
122+
"\n",
123+
"# Define loss and optimizer\n",
124+
"cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=pred, labels=y))\n",
125+
"optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate).minimize(cost)\n",
126+
"\n",
127+
"# Initializing the variables\n",
128+
"init = tf.global_variables_initializer()\n",
129+
"\n",
130+
"# Launch the graph\n",
131+
"with tf.Session() as sess:\n",
132+
" sess.run(init)\n",
133+
"\n",
134+
" # Training cycle\n",
135+
" for epoch in range(training_epochs):\n",
136+
" avg_cost = 0.\n",
137+
" total_batch = int(mnist.train.num_examples/batch_size)\n",
138+
" # Loop over all batches\n",
139+
" for i in range(total_batch):\n",
140+
" batch_x, batch_y = mnist.train.next_batch(batch_size)\n",
141+
" # Run optimization op (backprop) and cost op (to get loss value)\n",
142+
" _, c = sess.run([optimizer, cost], feed_dict={x: batch_x,\n",
143+
" y: batch_y})\n",
144+
" # Compute average loss\n",
145+
" avg_cost += c / total_batch\n",
146+
" # Display logs per epoch step\n",
147+
" if epoch % display_step == 0:\n",
148+
" print(\"Epoch:\", '%04d' % (epoch+1), \"cost=\", \\\n",
149+
" \"{:.9f}\".format(avg_cost))\n",
150+
" print(\"Optimization Finished!\")\n",
151+
"\n",
152+
" # Test model\n",
153+
" correct_prediction = tf.equal(tf.argmax(pred, 1), tf.argmax(y, 1))\n",
154+
" # Calculate accuracy\n",
155+
" accuracy = tf.reduce_mean(tf.cast(correct_prediction, \"float\"))\n",
156+
" print(\"Accuracy:\", accuracy.eval({x: mnist.test.images, y: mnist.test.labels}))\n"
157+
]
158+
},
159+
{
160+
"cell_type": "code",
161+
"execution_count": null,
162+
"metadata": {
163+
"collapsed": true
164+
},
165+
"outputs": [],
166+
"source": []
167+
}
168+
],
169+
"metadata": {
170+
"kernelspec": {
171+
"display_name": "Python 2",
172+
"language": "python",
173+
"name": "python2"
174+
},
175+
"language_info": {
176+
"codemirror_mode": {
177+
"name": "ipython",
178+
"version": 2
179+
},
180+
"file_extension": ".py",
181+
"mimetype": "text/x-python",
182+
"name": "python",
183+
"nbconvert_exporter": "python",
184+
"pygments_lexer": "ipython2",
185+
"version": "2.7.12"
186+
}
187+
},
188+
"nbformat": 4,
189+
"nbformat_minor": 2
190+
}

0 commit comments

Comments
 (0)