├── .gitignore
├── 021 Reading Data
├── random.data
└── 01 Reading Data From Files.ipynb
├── 001 Hello World
├── 01 Hello World.ipynb
└── 02 Numpy to TensorFlow.ipynb
├── 006 Matrix
└── 001 Matrix Multiplication.ipynb
├── 003 Basic Mathematical Operations
└── 001 Add.ipynb
├── 020 Data Manipulation
├── 05 range_input_producer.ipynb
├── 01 Convert To Tensor .ipynb
├── 03 Embedding Lookup.ipynb
├── 04 strided_slice.ipynb
└── 02 Split, Slice, and Join.ipynb
├── 005 Array Operations
├── 0001 One Hot Vector.ipynb
└── 002 ReduceMean.ipynb
├── 002 Variable
├── 02 Sharing Variables.ipynb
├── 01 Creation, Initialization, Saving, and Loading.ipynb
└── 03 Variable Scope.ipynb
├── 200 MNIST
└── 006 MNIST MultiLayer Perceptron.ipynb
├── 202 Logistic Regression with Regularization
├── titanic-logistic-regression-with-tensorflow.ipynb
└── test.csv
└── 010 Partial Differential Equations
└── 01 Partial Differential Equations.ipynb
/.gitignore:
--------------------------------------------------------------------------------
1 | # Project
2 | .idea
3 | .ipynb_checkpoints
4 |
5 | # Python
6 | *.pyc
7 | *.py~
8 |
9 | # Dataset
10 | _dataset
--------------------------------------------------------------------------------
/021 Reading Data/random.data:
--------------------------------------------------------------------------------
1 | 0,0.11074834488960074,1.7099860350765592,3.687902979662868,
2 | 1,-2.4967768565918362,1.4841636429678278,1.3211551538061077,
3 | 2,-0.4991637754688644,0.11133605098129183,5.089928448454691,
4 | 3,-0.8268802523654445,0.9344967881771236,5.439801533747579,
5 | 4,-0.9756282749554903,0.3363737199665942,3.291773330383719,
6 | 5,-2.8648743769196057,0.0661153649740855,5.03475588662393,
7 | 6,-0.3641233282871058,2.768433756668663,1.5993242291751433,
8 | 7,-0.26399830878806774,0.9900415796011218,2.1772022269390434,
9 | 8,-0.5667247601487606,0.5482186086310598,9.175617279771844,
10 | 9,-0.5170690934842015,2.2641959773355573,5.878481342892184,
11 |
--------------------------------------------------------------------------------
/001 Hello World/01 Hello World.ipynb:
--------------------------------------------------------------------------------
1 | {
2 | "cells": [
3 | {
4 | "cell_type": "code",
5 | "execution_count": 1,
6 | "metadata": {
7 | "collapsed": true
8 | },
9 | "outputs": [],
10 | "source": [
11 | "import tensorflow as tf\n",
12 | "from time import sleep"
13 | ]
14 | },
15 | {
16 | "cell_type": "code",
17 | "execution_count": 2,
18 | "metadata": {
19 | "collapsed": true
20 | },
21 | "outputs": [],
22 | "source": [
23 | "hello = tf.constant('Hello, TensorFlow and Anderson!')"
24 | ]
25 | },
26 | {
27 | "cell_type": "code",
28 | "execution_count": 3,
29 | "metadata": {
30 | "collapsed": false
31 | },
32 | "outputs": [
33 | {
34 | "name": "stdout",
35 | "output_type": "stream",
36 | "text": [
37 | "b'Hello, TensorFlow and Anderson!'\n"
38 | ]
39 | }
40 | ],
41 | "source": [
42 | "gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=0.01, allow_growth=True)\n",
43 | "with tf.Session(config=tf.ConfigProto(gpu_options=gpu_options)) as sess:\n",
44 | " print(sess.run(hello))"
45 | ]
46 | }
47 | ],
48 | "metadata": {
49 | "kernelspec": {
50 | "display_name": "Python 3",
51 | "language": "python",
52 | "name": "python3"
53 | },
54 | "language_info": {
55 | "codemirror_mode": {
56 | "name": "ipython",
57 | "version": 3.0
58 | },
59 | "file_extension": ".py",
60 | "mimetype": "text/x-python",
61 | "name": "python",
62 | "nbconvert_exporter": "python",
63 | "pygments_lexer": "ipython3",
64 | "version": "3.6.0"
65 | }
66 | },
67 | "nbformat": 4,
68 | "nbformat_minor": 0
69 | }
--------------------------------------------------------------------------------
/006 Matrix/001 Matrix Multiplication.ipynb:
--------------------------------------------------------------------------------
1 | {
2 | "cells": [
3 | {
4 | "cell_type": "code",
5 | "execution_count": 1,
6 | "metadata": {
7 | "collapsed": false
8 | },
9 | "outputs": [
10 | {
11 | "name": "stdout",
12 | "output_type": "stream",
13 | "text": [
14 | "Populating the interactive namespace from numpy and matplotlib\n"
15 | ]
16 | }
17 | ],
18 | "source": [
19 | "%pylab inline\n",
20 | "\n",
21 | "import tensorflow as tf\n",
22 | "import numpy as np"
23 | ]
24 | },
25 | {
26 | "cell_type": "code",
27 | "execution_count": 2,
28 | "metadata": {
29 | "collapsed": false
30 | },
31 | "outputs": [],
32 | "source": [
33 | "matrix1 = tf.constant([[1,2,3], [4,5,6]])\n",
34 | "matrix2 = tf.constant([[0, 1], [2, 2], [7, 9]])\n",
35 | "\n",
36 | "product_matrix = tf.matmul(matrix1, matrix2)"
37 | ]
38 | },
39 | {
40 | "cell_type": "code",
41 | "execution_count": 3,
42 | "metadata": {
43 | "collapsed": false
44 | },
45 | "outputs": [
46 | {
47 | "name": "stdout",
48 | "output_type": "stream",
49 | "text": [
50 | "[[25 32]\n",
51 | " [52 68]]\n"
52 | ]
53 | }
54 | ],
55 | "source": [
56 | "session = tf.Session()\n",
57 | "result = session.run(product_matrix)\n",
58 | "session.close()\n",
59 | "print result"
60 | ]
61 | }
62 | ],
63 | "metadata": {
64 | "kernelspec": {
65 | "display_name": "Python 2",
66 | "language": "python",
67 | "name": "python2"
68 | },
69 | "language_info": {
70 | "codemirror_mode": {
71 | "name": "ipython",
72 | "version": 2
73 | },
74 | "file_extension": ".py",
75 | "mimetype": "text/x-python",
76 | "name": "python",
77 | "nbconvert_exporter": "python",
78 | "pygments_lexer": "ipython2",
79 | "version": "2.7.10"
80 | }
81 | },
82 | "nbformat": 4,
83 | "nbformat_minor": 0
84 | }
85 |
--------------------------------------------------------------------------------
/001 Hello World/02 Numpy to TensorFlow.ipynb:
--------------------------------------------------------------------------------
1 | {
2 | "cells": [
3 | {
4 | "cell_type": "code",
5 | "execution_count": 1,
6 | "metadata": {
7 | "collapsed": true
8 | },
9 | "outputs": [],
10 | "source": [
11 | "import numpy as np\n",
12 | "import tensorflow as tf"
13 | ]
14 | },
15 | {
16 | "cell_type": "code",
17 | "execution_count": 2,
18 | "metadata": {
19 | "collapsed": true
20 | },
21 | "outputs": [],
22 | "source": [
23 | "gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=0.01, allow_growth=True)\n",
24 | "sess = tf.InteractiveSession(config=tf.ConfigProto(gpu_options=gpu_options))"
25 | ]
26 | },
27 | {
28 | "cell_type": "code",
29 | "execution_count": 14,
30 | "metadata": {
31 | "collapsed": false
32 | },
33 | "outputs": [
34 | {
35 | "name": "stdout",
36 | "output_type": "stream",
37 | "text": [
38 | "variable: [1 2 3]\n",
39 | "placeholder: [1 2 3]\n"
40 | ]
41 | }
42 | ],
43 | "source": [
44 | "np_value = np.array([1,2,3])\n",
45 | "\n",
46 | "variable = tf.Variable(np_value)\n",
47 | "placeholder = tf.placeholder('int32')\n",
48 | "\n",
49 | "sess.run(tf.global_variables_initializer())\n",
50 | "print('variable:', sess.run(variable))\n",
51 | "print('placeholder:', sess.run(placeholder, feed_dict={placeholder: np_value}))"
52 | ]
53 | }
54 | ],
55 | "metadata": {
56 | "kernelspec": {
57 | "display_name": "Python 3",
58 | "language": "python",
59 | "name": "python3"
60 | },
61 | "language_info": {
62 | "codemirror_mode": {
63 | "name": "ipython",
64 | "version": 3
65 | },
66 | "file_extension": ".py",
67 | "mimetype": "text/x-python",
68 | "name": "python",
69 | "nbconvert_exporter": "python",
70 | "pygments_lexer": "ipython3",
71 | "version": "3.6.0"
72 | }
73 | },
74 | "nbformat": 4,
75 | "nbformat_minor": 2
76 | }
77 |
--------------------------------------------------------------------------------
/003 Basic Mathematical Operations/001 Add.ipynb:
--------------------------------------------------------------------------------
1 | {
2 | "cells": [
3 | {
4 | "cell_type": "code",
5 | "execution_count": 1,
6 | "metadata": {
7 | "collapsed": true
8 | },
9 | "outputs": [],
10 | "source": [
11 | "import numpy as np\n",
12 | "import tensorflow as tf"
13 | ]
14 | },
15 | {
16 | "cell_type": "code",
17 | "execution_count": 16,
18 | "metadata": {
19 | "collapsed": true
20 | },
21 | "outputs": [],
22 | "source": [
23 | "X = tf.Variable([[1,2,3], [4,5,6], [7,8,9]])\n",
24 | "Y = tf.Variable(5)\n",
25 | "T = tf.Variable([0, 0, 1])\n",
26 | "\n",
27 | "s = tf.add(X, Y, name='sum')\n"
28 | ]
29 | },
30 | {
31 | "cell_type": "code",
32 | "execution_count": 21,
33 | "metadata": {
34 | "collapsed": false
35 | },
36 | "outputs": [
37 | {
38 | "name": "stdout",
39 | "output_type": "stream",
40 | "text": [
41 | "X + Y\n",
42 | "[[ 6 7 8]\n",
43 | " [ 9 10 11]\n",
44 | " [12 13 14]]\n",
45 | "\n",
46 | "X + T\n",
47 | "[[ 1 2 4]\n",
48 | " [ 4 5 7]\n",
49 | " [ 7 8 10]]\n"
50 | ]
51 | }
52 | ],
53 | "source": [
54 | "init_var_op = tf.initialize_all_variables()\n",
55 | "with tf.Session() as sess:\n",
56 | " sess.run(init_var_op)\n",
57 | " \n",
58 | " print 'X + Y' \n",
59 | " print sess.run( tf.add(X, Y, name='sum') )\n",
60 | " \n",
61 | " print \n",
62 | " print 'X + T'\n",
63 | " print sess.run(tf.add(X, T))"
64 | ]
65 | }
66 | ],
67 | "metadata": {
68 | "kernelspec": {
69 | "display_name": "Python 3",
70 | "language": "python",
71 | "name": "python3"
72 | },
73 | "language_info": {
74 | "codemirror_mode": {
75 | "name": "ipython",
76 | "version": 3
77 | },
78 | "file_extension": ".py",
79 | "mimetype": "text/x-python",
80 | "name": "python",
81 | "nbconvert_exporter": "python",
82 | "pygments_lexer": "ipython3",
83 | "version": "3.6.0"
84 | }
85 | },
86 | "nbformat": 4,
87 | "nbformat_minor": 0
88 | }
89 |
--------------------------------------------------------------------------------
/020 Data Manipulation/05 range_input_producer.ipynb:
--------------------------------------------------------------------------------
1 | {
2 | "cells": [
3 | {
4 | "cell_type": "code",
5 | "execution_count": 1,
6 | "metadata": {
7 | "collapsed": true
8 | },
9 | "outputs": [],
10 | "source": [
11 | "import tensorflow as tf"
12 | ]
13 | },
14 | {
15 | "cell_type": "markdown",
16 | "metadata": {},
17 | "source": [
18 | "### Create InteractiveSession"
19 | ]
20 | },
21 | {
22 | "cell_type": "code",
23 | "execution_count": 2,
24 | "metadata": {
25 | "collapsed": true
26 | },
27 | "outputs": [],
28 | "source": [
29 | "gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=0.001, allow_growth=True)\n",
30 | "sess = tf.InteractiveSession(config=tf.ConfigProto(gpu_options=gpu_options))\n",
31 | "init_op = tf.global_variables_initializer()\n",
32 | "sess.run(init_op)"
33 | ]
34 | },
35 | {
36 | "cell_type": "markdown",
37 | "metadata": {},
38 | "source": [
39 | "### Examples"
40 | ]
41 | },
42 | {
43 | "cell_type": "code",
44 | "execution_count": 3,
45 | "metadata": {
46 | "collapsed": false
47 | },
48 | "outputs": [
49 | {
50 | "data": {
51 | "text/plain": [
52 | "array([1, 2, 3, 4, 5, 6, 7, 8, 9], dtype=int32)"
53 | ]
54 | },
55 | "execution_count": 3,
56 | "metadata": {},
57 | "output_type": "execute_result"
58 | }
59 | ],
60 | "source": [
61 | "x = tf.range(1, 10, name='x')\n",
62 | "slice_end = tf.train.range_input_producer(1).dequeue()\n",
63 | "y = tf.slice(x, [0], [slice_end], name='y')\n",
64 | "sess.run(x)"
65 | ]
66 | }
67 | ],
68 | "metadata": {
69 | "kernelspec": {
70 | "display_name": "Python 3",
71 | "language": "python",
72 | "name": "python3"
73 | },
74 | "language_info": {
75 | "codemirror_mode": {
76 | "name": "ipython",
77 | "version": 3
78 | },
79 | "file_extension": ".py",
80 | "mimetype": "text/x-python",
81 | "name": "python",
82 | "nbconvert_exporter": "python",
83 | "pygments_lexer": "ipython3",
84 | "version": "3.6.0"
85 | }
86 | },
87 | "nbformat": 4,
88 | "nbformat_minor": 2
89 | }
90 |
--------------------------------------------------------------------------------
/005 Array Operations/0001 One Hot Vector.ipynb:
--------------------------------------------------------------------------------
1 | {
2 | "cells": [
3 | {
4 | "cell_type": "code",
5 | "execution_count": 7,
6 | "metadata": {
7 | "collapsed": true
8 | },
9 | "outputs": [],
10 | "source": [
11 | "import tensorflow as tf\n",
12 | "import tflearn\n",
13 | "import numpy as np"
14 | ]
15 | },
16 | {
17 | "cell_type": "markdown",
18 | "metadata": {},
19 | "source": [
20 | "### One Hot Vector\n",
21 | "\n",
22 | "data 변수값에는 다음과 같이 들어가 있습니다.\n",
23 | "```python\n",
24 | "data = [0, 1, 2, 3, 4.7, 4.1]\n",
25 | "```\n",
26 | "예제에서 one_hot() 함수는 data에 해당하는 index에 7을 넣고, 나머지는 0을 넣습니다.
\n",
27 | "중요하게 볼 부분은 4.7 이나 4.1 둘다 동일하게 4로 취급됩니다."
28 | ]
29 | },
30 | {
31 | "cell_type": "code",
32 | "execution_count": 34,
33 | "metadata": {
34 | "collapsed": false
35 | },
36 | "outputs": [],
37 | "source": [
38 | "x = tf.placeholder('int64', [None], name='x')\n",
39 | "x_one_hot = tf.one_hot(x, 10, 7., 0., name='x_one_hot')\n",
40 | "tf.one_hot?"
41 | ]
42 | },
43 | {
44 | "cell_type": "code",
45 | "execution_count": 44,
46 | "metadata": {
47 | "collapsed": false
48 | },
49 | "outputs": [
50 | {
51 | "name": "stdout",
52 | "output_type": "stream",
53 | "text": [
54 | "[[ 7. 0. 0. 0. 0. 0. 0. 0. 0. 0.]\n",
55 | " [ 0. 7. 0. 0. 0. 0. 0. 0. 0. 0.]\n",
56 | " [ 0. 0. 7. 0. 0. 0. 0. 0. 0. 0.]\n",
57 | " [ 0. 0. 0. 7. 0. 0. 0. 0. 0. 0.]\n",
58 | " [ 0. 0. 0. 0. 7. 0. 0. 0. 0. 0.]\n",
59 | " [ 0. 0. 0. 0. 7. 0. 0. 0. 0. 0.]]\n"
60 | ]
61 | }
62 | ],
63 | "source": [
64 | "with tf.Session() as sess:\n",
65 | " sess.run(tf.initialize_all_variables())\n",
66 | " \n",
67 | " data = [0, 1, 2, 3, 4.7, 4.1]\n",
68 | " print sess.run(x_one_hot, feed_dict={x: data})"
69 | ]
70 | }
71 | ],
72 | "metadata": {
73 | "kernelspec": {
74 | "display_name": "Python 2",
75 | "language": "python",
76 | "name": "python2"
77 | },
78 | "language_info": {
79 | "codemirror_mode": {
80 | "name": "ipython",
81 | "version": 2
82 | },
83 | "file_extension": ".py",
84 | "mimetype": "text/x-python",
85 | "name": "python",
86 | "nbconvert_exporter": "python",
87 | "pygments_lexer": "ipython2",
88 | "version": "2.7.12"
89 | }
90 | },
91 | "nbformat": 4,
92 | "nbformat_minor": 1
93 | }
94 |
--------------------------------------------------------------------------------
/002 Variable/02 Sharing Variables.ipynb:
--------------------------------------------------------------------------------
1 | {
2 | "cells": [
3 | {
4 | "cell_type": "markdown",
5 | "metadata": {},
6 | "source": [
7 | "# Sharing Variables\n",
8 | "\n",
9 | "하나의 Model에 관련된 Variables을 공유하기 위해서는 dictionary에 넣어서 다른곳에 쓰일경우 해당 dictionary자체를 공유합니다.
\n",
10 | "즉 Model에 있는 여러가지 variables각자 따로따로 관리하지 말고, dictionary에 다 넣을것. "
11 | ]
12 | },
13 | {
14 | "cell_type": "code",
15 | "execution_count": 1,
16 | "metadata": {
17 | "collapsed": true
18 | },
19 | "outputs": [],
20 | "source": [
21 | "import tensorflow as tf"
22 | ]
23 | },
24 | {
25 | "cell_type": "code",
26 | "execution_count": 5,
27 | "metadata": {
28 | "collapsed": false
29 | },
30 | "outputs": [],
31 | "source": [
32 | "input_images = tf.placeholder(\"float\", [None, 5, 32, 32])\n",
33 | "\n",
34 | "def create_model(input_images):\n",
35 | " conv1_weights = tf.Variable(tf.random_normal([5, 5, 32, 32]), name=\"conv1_weights\")\n",
36 | " conv1_biases = tf.Variable(tf.zeros([32]), name=\"conv1_biases\")\n",
37 | " conv1 = tf.nn.conv2d(input_images, conv1_weights, strides=[1, 1, 1, 1], padding='SAME')\n",
38 | " relu1 = tf.nn.relu(conv1 + conv1_biases)\n",
39 | "\n",
40 | " conv2_weights = tf.Variable(tf.random_normal([5, 5, 32, 32]), name=\"conv2_weights\")\n",
41 | " conv2_biases = tf.Variable(tf.zeros([32]), name=\"conv2_biases\")\n",
42 | " conv2 = tf.nn.conv2d(relu1, conv2_weights, strides=[1, 1, 1, 1], padding='SAME')\n",
43 | " \n",
44 | " model = {\n",
45 | " 'conv1_weights': conv1_weights, \n",
46 | " 'conv1_biases': conv1_biases, \n",
47 | " 'conv1': conv1,\n",
48 | " 'relu1': relu1,\n",
49 | " \n",
50 | " 'conv2_weights': conv2_weights,\n",
51 | " 'conv2_biases': conv2_biases,\n",
52 | " 'conv2': conv2\n",
53 | " }\n",
54 | " return model\n",
55 | "\n",
56 | "model = create_model(input_images)"
57 | ]
58 | }
59 | ],
60 | "metadata": {
61 | "kernelspec": {
62 | "display_name": "Python 3",
63 | "language": "python",
64 | "name": "python3"
65 | },
66 | "language_info": {
67 | "codemirror_mode": {
68 | "name": "ipython",
69 | "version": 3
70 | },
71 | "file_extension": ".py",
72 | "mimetype": "text/x-python",
73 | "name": "python",
74 | "nbconvert_exporter": "python",
75 | "pygments_lexer": "ipython3",
76 | "version": "3.6.0"
77 | }
78 | },
79 | "nbformat": 4,
80 | "nbformat_minor": 2
81 | }
82 |
--------------------------------------------------------------------------------
/020 Data Manipulation/01 Convert To Tensor .ipynb:
--------------------------------------------------------------------------------
1 | {
2 | "cells": [
3 | {
4 | "cell_type": "code",
5 | "execution_count": 2,
6 | "metadata": {
7 | "collapsed": true
8 | },
9 | "outputs": [],
10 | "source": [
11 | "import tensorflow as tf \n",
12 | "import numpy as np"
13 | ]
14 | },
15 | {
16 | "cell_type": "markdown",
17 | "metadata": {},
18 | "source": [
19 | "# Convert To Tensor \n",
20 | "\n",
21 | "tf.convert_to_tensor(value, dtype=None, name=None, preferred_dtype=None)"
22 | ]
23 | },
24 | {
25 | "cell_type": "markdown",
26 | "metadata": {},
27 | "source": [
28 | "### Example"
29 | ]
30 | },
31 | {
32 | "cell_type": "code",
33 | "execution_count": 6,
34 | "metadata": {
35 | "collapsed": false
36 | },
37 | "outputs": [],
38 | "source": [
39 | "def foo(arg):\n",
40 | " arg = tf.convert_to_tensor(arg, dtype='float32')\n",
41 | " return tf.matmul(arg, arg) + arg\n",
42 | "\n",
43 | "v1 = foo(tf.constant([[1.0, 2.0], [3.0, 4.0]]))\n",
44 | "v2 = foo([[1.0, 2.0], [3.0, 4.0]])\n",
45 | "v3 = foo(np.array([[1.0, 2.0], [3.0, 4.0]]))"
46 | ]
47 | },
48 | {
49 | "cell_type": "code",
50 | "execution_count": 12,
51 | "metadata": {
52 | "collapsed": false
53 | },
54 | "outputs": [
55 | {
56 | "name": "stdout",
57 | "output_type": "stream",
58 | "text": [
59 | "[[ 8. 12.]\n",
60 | " [ 18. 26.]]\n",
61 | "\n",
62 | "[[ 8. 12.]\n",
63 | " [ 18. 26.]]\n",
64 | "\n",
65 | "[[ 8. 12.]\n",
66 | " [ 18. 26.]]\n",
67 | "\n"
68 | ]
69 | }
70 | ],
71 | "source": [
72 | "gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=0.4, allow_growth=True)\n",
73 | "with tf.Session(config=tf.ConfigProto(gpu_options=gpu_options)) as sess:\n",
74 | " r1, r2, r3 = sess.run([v1, v2, v3])\n",
75 | " \n",
76 | " print(r1, end='\\n\\n')\n",
77 | " print(r2, end='\\n\\n')\n",
78 | " print(r3, end='\\n\\n')"
79 | ]
80 | }
81 | ],
82 | "metadata": {
83 | "kernelspec": {
84 | "display_name": "Python 3",
85 | "language": "python",
86 | "name": "python3"
87 | },
88 | "language_info": {
89 | "codemirror_mode": {
90 | "name": "ipython",
91 | "version": 3.0
92 | },
93 | "file_extension": ".py",
94 | "mimetype": "text/x-python",
95 | "name": "python",
96 | "nbconvert_exporter": "python",
97 | "pygments_lexer": "ipython3",
98 | "version": "3.6.0"
99 | }
100 | },
101 | "nbformat": 4,
102 | "nbformat_minor": 0
103 | }
--------------------------------------------------------------------------------
/020 Data Manipulation/03 Embedding Lookup.ipynb:
--------------------------------------------------------------------------------
1 | {
2 | "cells": [
3 | {
4 | "cell_type": "code",
5 | "execution_count": 1,
6 | "metadata": {
7 | "collapsed": true
8 | },
9 | "outputs": [],
10 | "source": [
11 | "import tensorflow as tf"
12 | ]
13 | },
14 | {
15 | "cell_type": "markdown",
16 | "metadata": {},
17 | "source": [
18 | "### Create a Session"
19 | ]
20 | },
21 | {
22 | "cell_type": "code",
23 | "execution_count": 6,
24 | "metadata": {
25 | "collapsed": true
26 | },
27 | "outputs": [],
28 | "source": [
29 | "init_op = tf.global_variables_initializer()\n",
30 | "\n",
31 | "gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=0.001, allow_growth=True)\n",
32 | "sess = tf.InteractiveSession(config=tf.ConfigProto(gpu_options=gpu_options))\n",
33 | "sess.run(init_op)"
34 | ]
35 | },
36 | {
37 | "cell_type": "markdown",
38 | "metadata": {},
39 | "source": [
40 | "### embedding_lookup\n",
41 | "\n",
42 | "params tensor에서 ids를 index해서 찾아낸 elements를 리턴합니다.
\n",
43 | "tf.gather와 유사합니다. \n",
44 | "\n",
45 | "또한 emgedding_lookup은 partition_strategy에 따라서 lookup 을 합니다. \n",
46 | "\n",
47 | "$$ \\begin{bmatrix} 100 & 200 \\\\ 300 & 400 \\end{bmatrix} $$\n",
48 | "\n",
49 | "에서 Index를 붙이면 다음과 같습니다. \n",
50 | "\n",
51 | "$$ \\begin{bmatrix} 0 & 2 \\\\ 1 & 3 \\end{bmatrix} $$"
52 | ]
53 | },
54 | {
55 | "cell_type": "code",
56 | "execution_count": 20,
57 | "metadata": {
58 | "collapsed": false
59 | },
60 | "outputs": [
61 | {
62 | "name": "stdout",
63 | "output_type": "stream",
64 | "text": [
65 | "embedding_lookup(params, ids):\t [10 10 30]\n",
66 | "tf.gather(params, ids): \t [10 10 30]\n"
67 | ]
68 | }
69 | ],
70 | "source": [
71 | "params = tf.constant([10, 20, 30, 40], dtype=tf.int32, name='params')\n",
72 | "ids = tf.constant([0, 0, 2])\n",
73 | "\n",
74 | "print('embedding_lookup(params, ids):\\t', tf.nn.embedding_lookup(params, ids).eval())\n",
75 | "print('tf.gather(params, ids): \\t', tf.gather(params, ids).eval())"
76 | ]
77 | },
78 | {
79 | "cell_type": "code",
80 | "execution_count": 48,
81 | "metadata": {
82 | "collapsed": false
83 | },
84 | "outputs": [
85 | {
86 | "name": "stdout",
87 | "output_type": "stream",
88 | "text": [
89 | "[100 300 200 400 100]\n"
90 | ]
91 | }
92 | ],
93 | "source": [
94 | "params1 = tf.constant([100,200])\n",
95 | "params2 = tf.constant([300,400])\n",
96 | "ids = tf.constant([0, 1, 2, 3, 0])\n",
97 | "print(tf.nn.embedding_lookup([params1, params2], ids).eval())"
98 | ]
99 | }
100 | ],
101 | "metadata": {
102 | "kernelspec": {
103 | "display_name": "Python 3",
104 | "language": "python",
105 | "name": "python3"
106 | },
107 | "language_info": {
108 | "codemirror_mode": {
109 | "name": "ipython",
110 | "version": 3
111 | },
112 | "file_extension": ".py",
113 | "mimetype": "text/x-python",
114 | "name": "python",
115 | "nbconvert_exporter": "python",
116 | "pygments_lexer": "ipython3",
117 | "version": "3.6.0"
118 | }
119 | },
120 | "nbformat": 4,
121 | "nbformat_minor": 2
122 | }
123 |
--------------------------------------------------------------------------------
/002 Variable/01 Creation, Initialization, Saving, and Loading.ipynb:
--------------------------------------------------------------------------------
1 | {
2 | "cells": [
3 | {
4 | "cell_type": "markdown",
5 | "metadata": {},
6 | "source": [
7 | "# Variables \n",
8 | "\n",
9 | "Variables은 Model을 training시에 parameters(weights)들을 update하거나 갖고 있을수 있습니다.
\n",
10 | "In-memory buffers이며 명시적으로 초기화시켜줘야 하며 저장하거나 불러올수 있습니다. "
11 | ]
12 | },
13 | {
14 | "cell_type": "code",
15 | "execution_count": 1,
16 | "metadata": {
17 | "collapsed": true
18 | },
19 | "outputs": [],
20 | "source": [
21 | "import tensorflow as tf"
22 | ]
23 | },
24 | {
25 | "cell_type": "markdown",
26 | "metadata": {},
27 | "source": [
28 | "### Creation\n",
29 | "\n",
30 | "Variables을 생성할때, Tensor를 초기값으로 전달해야 합니다. "
31 | ]
32 | },
33 | {
34 | "cell_type": "code",
35 | "execution_count": 2,
36 | "metadata": {
37 | "collapsed": true
38 | },
39 | "outputs": [],
40 | "source": [
41 | "weights = tf.Variable(tf.random_normal([784, 200], stddev=0.35), name='weights')\n",
42 | "biases = tf.Variable(tf.zeros([200]), name='biases')"
43 | ]
44 | },
45 | {
46 | "cell_type": "markdown",
47 | "metadata": {},
48 | "source": [
49 | "### Device Placement\n",
50 | "\n",
51 | "Variables는 특정 디바이스를 지정할수 있습니다.
\n",
52 | "하지만 v.assign() operators처럼 mutate하는 것들은 반드시 같은 device안에 존재해야 합니다. "
53 | ]
54 | },
55 | {
56 | "cell_type": "code",
57 | "execution_count": 3,
58 | "metadata": {
59 | "collapsed": true
60 | },
61 | "outputs": [],
62 | "source": [
63 | "with tf.device('/cpu:0'):\n",
64 | " cpu_v = tf.Variable(tf.random_normal([784, 200], stddev=0.5), name='cpu_vector')\n",
65 | " \n",
66 | "with tf.device('/gpu:0'):\n",
67 | " gpu_v = tf.Variable(tf.random_normal([784, 200], stddev=0.5), name='gpu_vector')"
68 | ]
69 | },
70 | {
71 | "cell_type": "markdown",
72 | "metadata": {},
73 | "source": [
74 | "### Initialization\n",
75 | "\n",
76 | "Variable initializers 는 반드시 명시적으로 실행되야 합니다. "
77 | ]
78 | },
79 | {
80 | "cell_type": "code",
81 | "execution_count": 7,
82 | "metadata": {
83 | "collapsed": false
84 | },
85 | "outputs": [],
86 | "source": [
87 | "init_op = tf.global_variables_initializer()\n",
88 | "\n",
89 | "gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=0.001, allow_growth=True)\n",
90 | "sess = tf.Session(config=tf.ConfigProto(gpu_options=gpu_options))\n",
91 | "sess.run(init_op)"
92 | ]
93 | },
94 | {
95 | "cell_type": "markdown",
96 | "metadata": {},
97 | "source": [
98 | "### Saving Variables"
99 | ]
100 | },
101 | {
102 | "cell_type": "code",
103 | "execution_count": 9,
104 | "metadata": {
105 | "collapsed": false
106 | },
107 | "outputs": [
108 | {
109 | "data": {
110 | "text/plain": [
111 | "'/tmp/variable_test.ckpt'"
112 | ]
113 | },
114 | "execution_count": 9,
115 | "metadata": {},
116 | "output_type": "execute_result"
117 | }
118 | ],
119 | "source": [
120 | "saver =tf.train.Saver()\n",
121 | "saver.save(sess, '/tmp/variable_test.ckpt')"
122 | ]
123 | },
124 | {
125 | "cell_type": "markdown",
126 | "metadata": {},
127 | "source": [
128 | "### Restoring Variables\n",
129 | "\n",
130 | "Restoring 할때는 initialization을 해줄 필요가 없습니다. "
131 | ]
132 | },
133 | {
134 | "cell_type": "code",
135 | "execution_count": 11,
136 | "metadata": {
137 | "collapsed": true
138 | },
139 | "outputs": [],
140 | "source": [
141 | "saver.restore(sess, '/tmp/variable_test.ckpt')"
142 | ]
143 | },
144 | {
145 | "cell_type": "markdown",
146 | "metadata": {},
147 | "source": [
148 | "### Variables to Numpy Array\n",
149 | "\n",
150 | "그냥 Run 실행시키면 됨"
151 | ]
152 | },
153 | {
154 | "cell_type": "code",
155 | "execution_count": 13,
156 | "metadata": {
157 | "collapsed": false
158 | },
159 | "outputs": [
160 | {
161 | "data": {
162 | "text/plain": [
163 | "array([[ 0.79427677, -0.70706618, -0.43262488, ..., 0.01068686,\n",
164 | " -0.085705 , -0.31455702],\n",
165 | " [-0.04460101, -0.02016782, -0.50548464, ..., -0.42810267,\n",
166 | " 0.20497903, -0.28702372],\n",
167 | " [-0.62877887, -0.51519108, 0.63525504, ..., 0.15575312,\n",
168 | " -0.09617887, 0.15659091],\n",
169 | " ..., \n",
170 | " [ 0.31423798, 0.09636419, -0.45797321, ..., 0.42629337,\n",
171 | " -0.85066891, 0.59309483],\n",
172 | " [ 0.19040468, 0.2698268 , -0.12512121, ..., 0.13218126,\n",
173 | " -0.6251471 , -0.35833758],\n",
174 | " [ 0.92575026, 0.0520521 , 0.03383748, ..., -0.25407696,\n",
175 | " -0.29910761, 0.25049213]], dtype=float32)"
176 | ]
177 | },
178 | "execution_count": 13,
179 | "metadata": {},
180 | "output_type": "execute_result"
181 | }
182 | ],
183 | "source": [
184 | "sess.run(weights)"
185 | ]
186 | }
187 | ],
188 | "metadata": {
189 | "kernelspec": {
190 | "display_name": "Python 3",
191 | "language": "python",
192 | "name": "python3"
193 | },
194 | "language_info": {
195 | "codemirror_mode": {
196 | "name": "ipython",
197 | "version": 3.0
198 | },
199 | "file_extension": ".py",
200 | "mimetype": "text/x-python",
201 | "name": "python",
202 | "nbconvert_exporter": "python",
203 | "pygments_lexer": "ipython3",
204 | "version": "3.6.0"
205 | }
206 | },
207 | "nbformat": 4,
208 | "nbformat_minor": 0
209 | }
--------------------------------------------------------------------------------
/002 Variable/03 Variable Scope.ipynb:
--------------------------------------------------------------------------------
1 | {
2 | "cells": [
3 | {
4 | "cell_type": "markdown",
5 | "metadata": {
6 | "deletable": true,
7 | "editable": true
8 | },
9 | "source": [
10 | "# Variable Scope\n",
11 | "\n",
12 | "Variable Scope mechanism은 2개의 함수로 구성되어 있습니다. \n",
13 | "\n",
14 | "* tf.get_variable(, , ): Creates or returns a variable with a given name.\n",
15 | "* tf.variable_scope(): Manages namespaces for names passed to tf.get_variable().\n",
16 | "\n",
17 | "tf.get_variable 의 경우, 없으면 만들기 때문에 direct로 특정 값을 보내지 않고, initializer를 passing 합니다.
\n",
18 | "initializers 는 예를들어 다음의 것들이 있습니다. \n",
19 | "\n",
20 | "* tf.constant_initializer(value) initializes everything to the provided value,\n",
21 | "* tf.random_uniform_initializer(a, b) initializes uniformly from [a, b],\n",
22 | "* tf.random_normal_initializer(mean, stddev) initializes from the normal distribution with the given mean and standard deviation."
23 | ]
24 | },
25 | {
26 | "cell_type": "markdown",
27 | "metadata": {
28 | "deletable": true,
29 | "editable": true
30 | },
31 | "source": [
32 | "### 예제 \n",
33 | "\n",
34 | "아래의 예제에서 convolution 2개를 만들고자 합니다.
\n",
35 | "이때 conv_relu에 사용된 이름들 'weights', 'biases', 등이 동일하기 때문에, conv_relu함수를 2번 사용할수가 없습니다.
\n",
36 | "이때 이름을 구분짓기 위해서 tf.variable_scope을 사용합니다. "
37 | ]
38 | },
39 | {
40 | "cell_type": "code",
41 | "execution_count": 1,
42 | "metadata": {
43 | "collapsed": true,
44 | "deletable": true,
45 | "editable": true
46 | },
47 | "outputs": [],
48 | "source": [
49 | "import tensorflow as tf"
50 | ]
51 | },
52 | {
53 | "cell_type": "code",
54 | "execution_count": 2,
55 | "metadata": {
56 | "collapsed": false,
57 | "deletable": true,
58 | "editable": true
59 | },
60 | "outputs": [],
61 | "source": [
62 | "input_images = tf.placeholder(\"float\", [None, 5, 32, 32])\n",
63 | "\n",
64 | "def conv_relu(input_layer, kernel_shape, bias_shape):\n",
65 | " weights = tf.get_variable('weights', kernel_shape, initializer=tf.random_normal_initializer())\n",
66 | " biases = tf.get_variable('biases', bias_shape, initializer=tf.constant_initializer(0.0))\n",
67 | " conv = tf.nn.conv2d(input_layer, weights, strides=[1,1,1,1], padding='SAME')\n",
68 | " return tf.nn.relu(conv + biases)\n",
69 | "\n",
70 | "def create_model(input_layer):\n",
71 | " with tf.variable_scope('conv1'):\n",
72 | " relu1 = conv_relu(input_layer, [5, 5, 32, 32], [32])\n",
73 | " \n",
74 | " with tf.variable_scope('conv2'):\n",
75 | " relu2 = conv_relu(relu1, [5, 5, 32, 32], [32])\n",
76 | " \n",
77 | " return relu2\n",
78 | "\n",
79 | "model = create_model(input_images)"
80 | ]
81 | },
82 | {
83 | "cell_type": "code",
84 | "execution_count": 3,
85 | "metadata": {
86 | "collapsed": false
87 | },
88 | "outputs": [
89 | {
90 | "data": {
91 | "text/plain": [
92 | "'conv2/Relu'"
93 | ]
94 | },
95 | "execution_count": 3,
96 | "metadata": {},
97 | "output_type": "execute_result"
98 | }
99 | ],
100 | "source": [
101 | "model.op.name"
102 | ]
103 | },
104 | {
105 | "cell_type": "markdown",
106 | "metadata": {
107 | "deletable": true,
108 | "editable": true
109 | },
110 | "source": [
111 | "하지만 model = create_model(input_images) 를 2번 실행시키면 ValueError Exception이 일어나게 됩니다.
\n",
112 | "tf.get_variable()은 이미 존재하는 variables들을 감지하고 실수로 공유되지 않도록 막습니다.
\n",
113 | "때문에 저런 에러가 일어난 것이고, share가 필요하다면 reuse_variables()를 사용합니다. "
114 | ]
115 | },
116 | {
117 | "cell_type": "code",
118 | "execution_count": 3,
119 | "metadata": {
120 | "collapsed": false,
121 | "deletable": true,
122 | "editable": true
123 | },
124 | "outputs": [],
125 | "source": [
126 | "with tf.variable_scope('image_filters') as scope:\n",
127 | " model2 = create_model(input_images)\n",
128 | " scope.reuse_variables()\n",
129 | " model3 = create_model(input_images)"
130 | ]
131 | },
132 | {
133 | "cell_type": "markdown",
134 | "metadata": {},
135 | "source": [
136 | "### Capturing Variable Scope\n",
137 | "\n",
138 | "scope자체를 공유할수도 있습니다. "
139 | ]
140 | },
141 | {
142 | "cell_type": "code",
143 | "execution_count": 4,
144 | "metadata": {
145 | "collapsed": false
146 | },
147 | "outputs": [],
148 | "source": [
149 | "with tf.variable_scope(\"foo\") as foo_scope:\n",
150 | " v = tf.get_variable(\"v\", [1])\n",
151 | "\n",
152 | "with tf.variable_scope(foo_scope):\n",
153 | " w = tf.get_variable(\"w\", [1])\n",
154 | "\n",
155 | "with tf.variable_scope(foo_scope, reuse=True):\n",
156 | " v1 = tf.get_variable(\"v\", [1])\n",
157 | " w1 = tf.get_variable(\"w\", [1])\n",
158 | " \n",
159 | "assert v1 is v\n",
160 | "assert w1 is w"
161 | ]
162 | },
163 | {
164 | "cell_type": "code",
165 | "execution_count": 5,
166 | "metadata": {
167 | "collapsed": false
168 | },
169 | "outputs": [
170 | {
171 | "name": "stdout",
172 | "output_type": "stream",
173 | "text": [
174 | "v.name: foo/v\n",
175 | "w.name: foo/w\n"
176 | ]
177 | }
178 | ],
179 | "source": [
180 | "print('v.name:', v.op.name)\n",
181 | "print('w.name:', w.op.name)"
182 | ]
183 | }
184 | ],
185 | "metadata": {
186 | "kernelspec": {
187 | "display_name": "Python 3",
188 | "language": "python",
189 | "name": "python3"
190 | },
191 | "language_info": {
192 | "codemirror_mode": {
193 | "name": "ipython",
194 | "version": 3
195 | },
196 | "file_extension": ".py",
197 | "mimetype": "text/x-python",
198 | "name": "python",
199 | "nbconvert_exporter": "python",
200 | "pygments_lexer": "ipython3",
201 | "version": "3.6.0"
202 | }
203 | },
204 | "nbformat": 4,
205 | "nbformat_minor": 2
206 | }
207 |
--------------------------------------------------------------------------------
/200 MNIST/006 MNIST MultiLayer Perceptron.ipynb:
--------------------------------------------------------------------------------
1 | {
2 | "cells": [
3 | {
4 | "cell_type": "code",
5 | "execution_count": 1,
6 | "metadata": {},
7 | "outputs": [
8 | {
9 | "name": "stdout",
10 | "output_type": "stream",
11 | "text": [
12 | "Extracting /tmp/data/train-images-idx3-ubyte.gz\n",
13 | "Extracting /tmp/data/train-labels-idx1-ubyte.gz\n",
14 | "Extracting /tmp/data/t10k-images-idx3-ubyte.gz\n",
15 | "Extracting /tmp/data/t10k-labels-idx1-ubyte.gz\n"
16 | ]
17 | }
18 | ],
19 | "source": [
20 | "# Import MINST data\n",
21 | "from tensorflow.examples.tutorials.mnist import input_data\n",
22 | "mnist = input_data.read_data_sets(\"/tmp/data/\", one_hot=True)\n",
23 | "\n",
24 | "import tensorflow as tf"
25 | ]
26 | },
27 | {
28 | "cell_type": "code",
29 | "execution_count": 2,
30 | "metadata": {},
31 | "outputs": [],
32 | "source": [
33 | "# Parameters\n",
34 | "learning_rate = 0.001\n",
35 | "training_epochs = 15\n",
36 | "batch_size = 100\n",
37 | "display_step = 1\n",
38 | "\n",
39 | "# Network Parameters\n",
40 | "n_hidden_1 = 256 # 1st layer number of features\n",
41 | "n_hidden_2 = 256 # 2nd layer number of features\n",
42 | "n_input = 784 # MNIST data input (img shape: 28*28)\n",
43 | "n_classes = 10 # MNIST total classes (0-9 digits)\n",
44 | "\n",
45 | "# tf Graph input\n",
46 | "x = tf.placeholder(\"float\", [None, n_input])\n",
47 | "y = tf.placeholder(\"float\", [None, n_classes])"
48 | ]
49 | },
50 | {
51 | "cell_type": "code",
52 | "execution_count": 3,
53 | "metadata": {},
54 | "outputs": [],
55 | "source": [
56 | "# Create model\n",
57 | "def multilayer_perceptron(x, weights, biases):\n",
58 | " # Hidden layer with RELU activation\n",
59 | " layer_1 = tf.add(tf.matmul(x, weights['h1']), biases['b1'])\n",
60 | " layer_1 = tf.nn.relu(layer_1)\n",
61 | " # Hidden layer with RELU activation\n",
62 | " layer_2 = tf.add(tf.matmul(layer_1, weights['h2']), biases['b2'])\n",
63 | " layer_2 = tf.nn.relu(layer_2)\n",
64 | " # Output layer with linear activation\n",
65 | " out_layer = tf.matmul(layer_2, weights['out']) + biases['out']\n",
66 | " return out_layer"
67 | ]
68 | },
69 | {
70 | "cell_type": "code",
71 | "execution_count": 4,
72 | "metadata": {},
73 | "outputs": [],
74 | "source": [
75 | "# Store layers weight & bias\n",
76 | "weights = {\n",
77 | " 'h1': tf.Variable(tf.random_normal([n_input, n_hidden_1])),\n",
78 | " 'h2': tf.Variable(tf.random_normal([n_hidden_1, n_hidden_2])),\n",
79 | " 'out': tf.Variable(tf.random_normal([n_hidden_2, n_classes]))\n",
80 | "}\n",
81 | "biases = {\n",
82 | " 'b1': tf.Variable(tf.random_normal([n_hidden_1])),\n",
83 | " 'b2': tf.Variable(tf.random_normal([n_hidden_2])),\n",
84 | " 'out': tf.Variable(tf.random_normal([n_classes]))\n",
85 | "}\n",
86 | "\n",
87 | "# Construct model\n",
88 | "pred = multilayer_perceptron(x, weights, biases)\n",
89 | "\n",
90 | "# Define loss and optimizer\n",
91 | "cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y, logits=pred))\n",
92 | "optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate).minimize(cost)\n",
93 | "\n",
94 | "# Initializing the variables\n",
95 | "init = tf.global_variables_initializer()"
96 | ]
97 | },
98 | {
99 | "cell_type": "code",
100 | "execution_count": 5,
101 | "metadata": {},
102 | "outputs": [
103 | {
104 | "name": "stdout",
105 | "output_type": "stream",
106 | "text": [
107 | "Epoch: 0001 cost= 199.156035394\n",
108 | "Epoch: 0002 cost= 43.510258431\n",
109 | "Epoch: 0003 cost= 26.776377823\n",
110 | "Epoch: 0004 cost= 18.127787209\n",
111 | "Epoch: 0005 cost= 13.059884647\n",
112 | "Epoch: 0006 cost= 9.537693703\n",
113 | "Epoch: 0007 cost= 6.947895438\n",
114 | "Epoch: 0008 cost= 5.192850764\n",
115 | "Epoch: 0009 cost= 3.830189170\n",
116 | "Epoch: 0010 cost= 2.755372774\n",
117 | "Epoch: 0011 cost= 1.976166025\n",
118 | "Epoch: 0012 cost= 1.491869799\n",
119 | "Epoch: 0013 cost= 1.114585339\n",
120 | "Epoch: 0014 cost= 0.874315661\n",
121 | "Epoch: 0015 cost= 0.677450839\n",
122 | "batch_y: \n",
123 | "Optimization Finished!\n",
124 | "Accuracy: 0.9463\n"
125 | ]
126 | }
127 | ],
128 | "source": [
129 | "# Launch the graph\n",
130 | "with tf.Session() as sess:\n",
131 | " sess.run(init)\n",
132 | "\n",
133 | " # Training cycle\n",
134 | " for epoch in range(training_epochs):\n",
135 | " avg_cost = 0.\n",
136 | " total_batch = int(mnist.train.num_examples/batch_size)\n",
137 | " # Loop over all batches\n",
138 | " for i in range(total_batch):\n",
139 | " batch_x, batch_y = mnist.train.next_batch(batch_size)\n",
140 | " # Run optimization op (backprop) and cost op (to get loss value)\n",
141 | " \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 'batch_y: '\n",
151 | " print \"Optimization Finished!\"\n",
152 | "\n",
153 | " # Test model\n",
154 | " correct_prediction = tf.equal(tf.argmax(pred, 1), tf.argmax(y, 1))\n",
155 | " # Calculate accuracy\n",
156 | " accuracy = tf.reduce_mean(tf.cast(correct_prediction, \"float\"))\n",
157 | " print \"Accuracy:\", accuracy.eval({x: mnist.test.images, y: mnist.test.labels})"
158 | ]
159 | }
160 | ],
161 | "metadata": {
162 | "kernelspec": {
163 | "display_name": "Python 3",
164 | "language": "python",
165 | "name": "python3"
166 | },
167 | "language_info": {
168 | "codemirror_mode": {
169 | "name": "ipython",
170 | "version": 3
171 | },
172 | "file_extension": ".py",
173 | "mimetype": "text/x-python",
174 | "name": "python",
175 | "nbconvert_exporter": "python",
176 | "pygments_lexer": "ipython3",
177 | "version": "3.6.1"
178 | }
179 | },
180 | "nbformat": 4,
181 | "nbformat_minor": 1
182 | }
183 |
--------------------------------------------------------------------------------
/005 Array Operations/002 ReduceMean.ipynb:
--------------------------------------------------------------------------------
1 | {
2 | "cells": [
3 | {
4 | "cell_type": "code",
5 | "execution_count": 1,
6 | "metadata": {
7 | "collapsed": true
8 | },
9 | "outputs": [],
10 | "source": [
11 | "import tensorflow as tf\n",
12 | "import numpy as np"
13 | ]
14 | },
15 | {
16 | "cell_type": "code",
17 | "execution_count": 4,
18 | "metadata": {
19 | "collapsed": true
20 | },
21 | "outputs": [],
22 | "source": [
23 | "init_op = tf.global_variables_initializer()\n",
24 | "gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=0.001, allow_growth=True)\n",
25 | "sess = tf.InteractiveSession(config=tf.ConfigProto(gpu_options=gpu_options))\n",
26 | "sess.run(init_op)"
27 | ]
28 | },
29 | {
30 | "cell_type": "code",
31 | "execution_count": 3,
32 | "metadata": {
33 | "collapsed": false
34 | },
35 | "outputs": [
36 | {
37 | "data": {
38 | "text/plain": [
39 | "(3, 1, 4)"
40 | ]
41 | },
42 | "execution_count": 3,
43 | "metadata": {},
44 | "output_type": "execute_result"
45 | }
46 | ],
47 | "source": [
48 | "data = np.array([[[1, 2, 3, 4]], \n",
49 | " [[5, 6, 7, 8]], \n",
50 | " [[0, 10, 20, 30]]])"
51 | ]
52 | },
53 | {
54 | "cell_type": "code",
55 | "execution_count": 20,
56 | "metadata": {
57 | "collapsed": false
58 | },
59 | "outputs": [
60 | {
61 | "data": {
62 | "text/plain": [
63 | "[8.0,\n",
64 | " array([[ 2., 6., 10., 14.]], dtype=float32),\n",
65 | " array([[ 1., 2., 3., 4.],\n",
66 | " [ 5., 6., 7., 8.],\n",
67 | " [ 0., 10., 20., 30.]], dtype=float32),\n",
68 | " array([[ 2.5],\n",
69 | " [ 6.5],\n",
70 | " [ 15. ]], dtype=float32),\n",
71 | " array([ 2.5, 6.5, 15. ], dtype=float32)]"
72 | ]
73 | },
74 | "execution_count": 20,
75 | "metadata": {},
76 | "output_type": "execute_result"
77 | }
78 | ],
79 | "source": [
80 | "inputs = tf.placeholder('float32', shape=(None, None, 4))\n",
81 | "mean1 = tf.reduce_mean(inputs)\n",
82 | "mean2 = tf.reduce_mean(inputs, reduction_indices=[0])\n",
83 | "mean3 = tf.reduce_mean(inputs, reduction_indices=[1])\n",
84 | "mean4 = tf.reduce_mean(inputs, reduction_indices=[2])\n",
85 | "mean5 = tf.reduce_mean(inputs, reduction_indices=[1, 2])\n",
86 | "\n",
87 | "init_op = tf.global_variables_initializer()\n",
88 | "sess.run(init_op)\n",
89 | "\n",
90 | "sess.run([mean1, mean2, mean3, mean4, mean5], feed_dict={inputs: data})"
91 | ]
92 | },
93 | {
94 | "cell_type": "markdown",
95 | "metadata": {},
96 | "source": [
97 | "## reduce_mean(inputs)\n",
98 | "\n",
99 | "그냥 reduce_mean 사용시, 전체 elements들의 평균을 구합니다.
\n",
100 | "다음과 동일합니다. \n",
101 | "$$ \\frac{1+2+3+4+5+6+7+8+10+20+30}{12} $$"
102 | ]
103 | },
104 | {
105 | "cell_type": "code",
106 | "execution_count": 13,
107 | "metadata": {
108 | "collapsed": false
109 | },
110 | "outputs": [
111 | {
112 | "data": {
113 | "text/plain": [
114 | "8.0"
115 | ]
116 | },
117 | "execution_count": 13,
118 | "metadata": {},
119 | "output_type": "execute_result"
120 | }
121 | ],
122 | "source": [
123 | "data.mean()"
124 | ]
125 | },
126 | {
127 | "cell_type": "code",
128 | "execution_count": 14,
129 | "metadata": {
130 | "collapsed": false
131 | },
132 | "outputs": [
133 | {
134 | "data": {
135 | "text/plain": [
136 | "8.0"
137 | ]
138 | },
139 | "execution_count": 14,
140 | "metadata": {},
141 | "output_type": "execute_result"
142 | }
143 | ],
144 | "source": [
145 | "sess.run(mean1, feed_dict={inputs: data})"
146 | ]
147 | },
148 | {
149 | "cell_type": "markdown",
150 | "metadata": {},
151 | "source": [
152 | "## reduce_mean(inputs, reduction_indices=[0])\n",
153 | "\n",
154 | "columns들의 평균을 구합니다. "
155 | ]
156 | },
157 | {
158 | "cell_type": "code",
159 | "execution_count": 15,
160 | "metadata": {
161 | "collapsed": false
162 | },
163 | "outputs": [
164 | {
165 | "data": {
166 | "text/plain": [
167 | "array([[ 2., 6., 10., 14.]])"
168 | ]
169 | },
170 | "execution_count": 15,
171 | "metadata": {},
172 | "output_type": "execute_result"
173 | }
174 | ],
175 | "source": [
176 | "data.mean(axis=0)"
177 | ]
178 | },
179 | {
180 | "cell_type": "code",
181 | "execution_count": 16,
182 | "metadata": {
183 | "collapsed": false
184 | },
185 | "outputs": [
186 | {
187 | "data": {
188 | "text/plain": [
189 | "array([[ 2., 6., 10., 14.]], dtype=float32)"
190 | ]
191 | },
192 | "execution_count": 16,
193 | "metadata": {},
194 | "output_type": "execute_result"
195 | }
196 | ],
197 | "source": [
198 | "sess.run(mean2, feed_dict={inputs: data})"
199 | ]
200 | },
201 | {
202 | "cell_type": "markdown",
203 | "metadata": {},
204 | "source": [
205 | "## reduce_mean(inputs, reduction_indices=[2])\n",
206 | "\n",
207 | "rows의 평균을 구합니다."
208 | ]
209 | },
210 | {
211 | "cell_type": "code",
212 | "execution_count": 17,
213 | "metadata": {
214 | "collapsed": false
215 | },
216 | "outputs": [
217 | {
218 | "data": {
219 | "text/plain": [
220 | "array([[ 2.5],\n",
221 | " [ 6.5],\n",
222 | " [ 15. ]])"
223 | ]
224 | },
225 | "execution_count": 17,
226 | "metadata": {},
227 | "output_type": "execute_result"
228 | }
229 | ],
230 | "source": [
231 | "data.mean(axis=2)"
232 | ]
233 | },
234 | {
235 | "cell_type": "code",
236 | "execution_count": 18,
237 | "metadata": {
238 | "collapsed": false
239 | },
240 | "outputs": [
241 | {
242 | "data": {
243 | "text/plain": [
244 | "array([[ 2.5],\n",
245 | " [ 6.5],\n",
246 | " [ 15. ]], dtype=float32)"
247 | ]
248 | },
249 | "execution_count": 18,
250 | "metadata": {},
251 | "output_type": "execute_result"
252 | }
253 | ],
254 | "source": [
255 | "sess.run(mean4, feed_dict={inputs: data})"
256 | ]
257 | },
258 | {
259 | "cell_type": "markdown",
260 | "metadata": {},
261 | "source": [
262 | "## reduce_mean(inputs, reduction_indices=[1, 2])"
263 | ]
264 | },
265 | {
266 | "cell_type": "code",
267 | "execution_count": 25,
268 | "metadata": {
269 | "collapsed": false
270 | },
271 | "outputs": [
272 | {
273 | "data": {
274 | "text/plain": [
275 | "array([ 2.5, 6.5, 15. ])"
276 | ]
277 | },
278 | "execution_count": 25,
279 | "metadata": {},
280 | "output_type": "execute_result"
281 | }
282 | ],
283 | "source": [
284 | "data.mean(axis=2).mean(axis=1)"
285 | ]
286 | },
287 | {
288 | "cell_type": "code",
289 | "execution_count": 26,
290 | "metadata": {
291 | "collapsed": false
292 | },
293 | "outputs": [
294 | {
295 | "data": {
296 | "text/plain": [
297 | "array([ 2.5, 6.5, 15. ], dtype=float32)"
298 | ]
299 | },
300 | "execution_count": 26,
301 | "metadata": {},
302 | "output_type": "execute_result"
303 | }
304 | ],
305 | "source": [
306 | "sess.run(mean5, feed_dict={inputs: data})"
307 | ]
308 | }
309 | ],
310 | "metadata": {
311 | "kernelspec": {
312 | "display_name": "Python 3",
313 | "language": "python",
314 | "name": "python3"
315 | },
316 | "language_info": {
317 | "codemirror_mode": {
318 | "name": "ipython",
319 | "version": 3
320 | },
321 | "file_extension": ".py",
322 | "mimetype": "text/x-python",
323 | "name": "python",
324 | "nbconvert_exporter": "python",
325 | "pygments_lexer": "ipython3",
326 | "version": "3.6.0"
327 | }
328 | },
329 | "nbformat": 4,
330 | "nbformat_minor": 2
331 | }
332 |
--------------------------------------------------------------------------------
/020 Data Manipulation/04 strided_slice.ipynb:
--------------------------------------------------------------------------------
1 | {
2 | "cells": [
3 | {
4 | "cell_type": "markdown",
5 | "metadata": {},
6 | "source": [
7 | "# strided_slice\n",
8 | "\n",
9 | "```\n",
10 | "tf.strided_slice(input_, begin, end, strides=None, begin_mask=0, end_mask=0, ellipsis_mask=0, new_axis_mask=0, shrink_axis_mask=0, var=None, name=None)\n",
11 | "```\n",
12 | "\n",
13 | "begin과 end사이의 데이터를 extract합니다."
14 | ]
15 | },
16 | {
17 | "cell_type": "code",
18 | "execution_count": 2,
19 | "metadata": {
20 | "collapsed": true
21 | },
22 | "outputs": [],
23 | "source": [
24 | "import tensorflow as tf\n",
25 | "import numpy as np"
26 | ]
27 | },
28 | {
29 | "cell_type": "markdown",
30 | "metadata": {},
31 | "source": [
32 | "### Create InteractiveSession"
33 | ]
34 | },
35 | {
36 | "cell_type": "code",
37 | "execution_count": 5,
38 | "metadata": {
39 | "collapsed": true
40 | },
41 | "outputs": [],
42 | "source": [
43 | "gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=0.001, allow_growth=True)\n",
44 | "sess = tf.InteractiveSession(config=tf.ConfigProto(gpu_options=gpu_options))\n",
45 | "init_op = tf.global_variables_initializer()\n",
46 | "sess.run(init_op)"
47 | ]
48 | },
49 | {
50 | "cell_type": "markdown",
51 | "metadata": {},
52 | "source": [
53 | "### Data"
54 | ]
55 | },
56 | {
57 | "cell_type": "code",
58 | "execution_count": 25,
59 | "metadata": {
60 | "collapsed": false
61 | },
62 | "outputs": [
63 | {
64 | "data": {
65 | "text/plain": [
66 | "array([[[ 1.1, 1.2, 1.3],\n",
67 | " [ 2.1, 2.2, 2.3]],\n",
68 | "\n",
69 | " [[ 3.1, 3.2, 3.3],\n",
70 | " [ 4.1, 4.2, 4.3]],\n",
71 | "\n",
72 | " [[ 5.1, 5.2, 5.3],\n",
73 | " [ 6.1, 6.2, 6.3]]])"
74 | ]
75 | },
76 | "execution_count": 25,
77 | "metadata": {},
78 | "output_type": "execute_result"
79 | }
80 | ],
81 | "source": [
82 | "data = np.array([[[1.1, 1.2, 1.3], [2.1, 2.2, 2.3]], \n",
83 | " [[3.1, 3.2, 3.3], [4.1, 4.2, 4.3]], \n",
84 | " [[5.1, 5.2, 5.3], [6.1, 6.2, 6.3]]])\n",
85 | "inputs = tf.convert_to_tensor(data)\n",
86 | "sess.run(inputs)"
87 | ]
88 | },
89 | {
90 | "cell_type": "markdown",
91 | "metadata": {},
92 | "source": [
93 | "### Examples (using begin and end)"
94 | ]
95 | },
96 | {
97 | "cell_type": "code",
98 | "execution_count": 26,
99 | "metadata": {
100 | "collapsed": false
101 | },
102 | "outputs": [
103 | {
104 | "data": {
105 | "text/plain": [
106 | "array([[[ 1.1, 1.2, 1.3],\n",
107 | " [ 2.1, 2.2, 2.3]],\n",
108 | "\n",
109 | " [[ 3.1, 3.2, 3.3],\n",
110 | " [ 4.1, 4.2, 4.3]],\n",
111 | "\n",
112 | " [[ 5.1, 5.2, 5.3],\n",
113 | " [ 6.1, 6.2, 6.3]]])"
114 | ]
115 | },
116 | "execution_count": 26,
117 | "metadata": {},
118 | "output_type": "execute_result"
119 | }
120 | ],
121 | "source": [
122 | "sliced1 = tf.strided_slice(inputs, begin=[0], end=[3])\n",
123 | "sess.run(sliced1)"
124 | ]
125 | },
126 | {
127 | "cell_type": "code",
128 | "execution_count": 38,
129 | "metadata": {
130 | "collapsed": false
131 | },
132 | "outputs": [
133 | {
134 | "data": {
135 | "text/plain": [
136 | "array([[[ 3.1, 3.2, 3.3]],\n",
137 | "\n",
138 | " [[ 5.1, 5.2, 5.3]]])"
139 | ]
140 | },
141 | "execution_count": 38,
142 | "metadata": {},
143 | "output_type": "execute_result"
144 | }
145 | ],
146 | "source": [
147 | "sliced1 = tf.strided_slice(inputs, begin=[1, 0], end=[3, 1])\n",
148 | "sess.run(sliced1)"
149 | ]
150 | },
151 | {
152 | "cell_type": "code",
153 | "execution_count": 45,
154 | "metadata": {
155 | "collapsed": false
156 | },
157 | "outputs": [
158 | {
159 | "data": {
160 | "text/plain": [
161 | "array([[[ 3.1, 3.2, 3.3],\n",
162 | " [ 4.1, 4.2, 4.3]],\n",
163 | "\n",
164 | " [[ 5.1, 5.2, 5.3],\n",
165 | " [ 6.1, 6.2, 6.3]]])"
166 | ]
167 | },
168 | "execution_count": 45,
169 | "metadata": {},
170 | "output_type": "execute_result"
171 | }
172 | ],
173 | "source": [
174 | "sliced1 = tf.strided_slice(inputs, begin=[1, 0], end=[3, 2])\n",
175 | "sess.run(sliced1)"
176 | ]
177 | },
178 | {
179 | "cell_type": "code",
180 | "execution_count": 46,
181 | "metadata": {
182 | "collapsed": false
183 | },
184 | "outputs": [
185 | {
186 | "data": {
187 | "text/plain": [
188 | "array([[[ 3.1]],\n",
189 | "\n",
190 | " [[ 5.1]]])"
191 | ]
192 | },
193 | "execution_count": 46,
194 | "metadata": {},
195 | "output_type": "execute_result"
196 | }
197 | ],
198 | "source": [
199 | "sliced1 = tf.strided_slice(inputs, begin=[1, 0, 0], end=[3, 1, 1])\n",
200 | "sess.run(sliced1)"
201 | ]
202 | },
203 | {
204 | "cell_type": "code",
205 | "execution_count": 61,
206 | "metadata": {
207 | "collapsed": false
208 | },
209 | "outputs": [
210 | {
211 | "data": {
212 | "text/plain": [
213 | "array([[[ 1.1, 1.2]],\n",
214 | "\n",
215 | " [[ 3.1, 3.2]],\n",
216 | "\n",
217 | " [[ 5.1, 5.2]]])"
218 | ]
219 | },
220 | "execution_count": 61,
221 | "metadata": {},
222 | "output_type": "execute_result"
223 | }
224 | ],
225 | "source": [
226 | "sliced1 = tf.strided_slice(inputs, begin=[0, 0, 0], end=[3, 1, -1])\n",
227 | "sess.run(sliced1)"
228 | ]
229 | },
230 | {
231 | "cell_type": "code",
232 | "execution_count": 64,
233 | "metadata": {
234 | "collapsed": false
235 | },
236 | "outputs": [
237 | {
238 | "data": {
239 | "text/plain": [
240 | "array([[[ 1.1]],\n",
241 | "\n",
242 | " [[ 3.1]]])"
243 | ]
244 | },
245 | "execution_count": 64,
246 | "metadata": {},
247 | "output_type": "execute_result"
248 | }
249 | ],
250 | "source": [
251 | "sliced1 = tf.strided_slice(inputs, begin=[0, 0, 0], end=[-1, 1, 1])\n",
252 | "sess.run(sliced1)"
253 | ]
254 | },
255 | {
256 | "cell_type": "markdown",
257 | "metadata": {},
258 | "source": [
259 | "### Examples (using strides)\n",
260 | "\n",
261 | "strides 값이 모두 1값을 갖은 vector라면.. 안쓴거와 동일합니다."
262 | ]
263 | },
264 | {
265 | "cell_type": "code",
266 | "execution_count": 94,
267 | "metadata": {
268 | "collapsed": false
269 | },
270 | "outputs": [
271 | {
272 | "name": "stdout",
273 | "output_type": "stream",
274 | "text": [
275 | "no strides\n",
276 | "[[[ 1.1 1.2 1.3]]\n",
277 | "\n",
278 | " [[ 3.1 3.2 3.3]]]\n",
279 | "\n",
280 | "stride used\n",
281 | "[[[ 1.1 1.2 1.3]]\n",
282 | "\n",
283 | " [[ 3.1 3.2 3.3]]]\n"
284 | ]
285 | }
286 | ],
287 | "source": [
288 | "sliced1 = tf.strided_slice(inputs, begin=[0, 0, 0], end=[2, 1, 3])\n",
289 | "print('no strides')\n",
290 | "print(sess.run(sliced1))\n",
291 | "\n",
292 | "sliced1 = tf.strided_slice(inputs, begin=[0, 0, 0], end=[2, 1, 3], strides=[1, 1, 1])\n",
293 | "print('\\nstride used')\n",
294 | "print(sess.run(sliced1))"
295 | ]
296 | },
297 | {
298 | "cell_type": "code",
299 | "execution_count": 108,
300 | "metadata": {
301 | "collapsed": false
302 | },
303 | "outputs": [
304 | {
305 | "name": "stdout",
306 | "output_type": "stream",
307 | "text": [
308 | "[[[ 1.1 1.3]]\n",
309 | "\n",
310 | " [[ 5.1 5.3]]]\n"
311 | ]
312 | }
313 | ],
314 | "source": [
315 | "sliced1 = tf.strided_slice(inputs, begin=[0, 0, 0], end=[3, 1, 3], strides=[2, 1, 2])\n",
316 | "print(sess.run(sliced1))"
317 | ]
318 | }
319 | ],
320 | "metadata": {
321 | "kernelspec": {
322 | "display_name": "Python 3",
323 | "language": "python",
324 | "name": "python3"
325 | },
326 | "language_info": {
327 | "codemirror_mode": {
328 | "name": "ipython",
329 | "version": 3
330 | },
331 | "file_extension": ".py",
332 | "mimetype": "text/x-python",
333 | "name": "python",
334 | "nbconvert_exporter": "python",
335 | "pygments_lexer": "ipython3",
336 | "version": "3.6.0"
337 | }
338 | },
339 | "nbformat": 4,
340 | "nbformat_minor": 2
341 | }
342 |
--------------------------------------------------------------------------------
/020 Data Manipulation/02 Split, Slice, and Join.ipynb:
--------------------------------------------------------------------------------
1 | {
2 | "cells": [
3 | {
4 | "cell_type": "code",
5 | "execution_count": 80,
6 | "metadata": {
7 | "collapsed": true
8 | },
9 | "outputs": [],
10 | "source": [
11 | "import tensorflow as tf\n",
12 | "import numpy as np\n",
13 | "\n",
14 | "from pprint import pprint as pp"
15 | ]
16 | },
17 | {
18 | "cell_type": "markdown",
19 | "metadata": {},
20 | "source": [
21 | "## Create a Interactive Session"
22 | ]
23 | },
24 | {
25 | "cell_type": "code",
26 | "execution_count": 7,
27 | "metadata": {
28 | "collapsed": true
29 | },
30 | "outputs": [],
31 | "source": [
32 | "gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=0.01, allow_growth=True)\n",
33 | "sess = tf.InteractiveSession(config=tf.ConfigProto(gpu_options=gpu_options))"
34 | ]
35 | },
36 | {
37 | "cell_type": "markdown",
38 | "metadata": {},
39 | "source": [
40 | "## tf.slice(input_, begin, size, name=None)\n",
41 | "\n",
42 | "begin vector 가 예를 들어서 [0, 0, 1] 이고, size vector가 [1, 2, 3] 이라면 다음과 같은 slice을 하게 됩니다. \n",
43 | "\n",
44 | "> data[0:0+1, 0:0+2, 1:1+3]"
45 | ]
46 | },
47 | {
48 | "cell_type": "code",
49 | "execution_count": 59,
50 | "metadata": {
51 | "collapsed": false
52 | },
53 | "outputs": [
54 | {
55 | "name": "stdout",
56 | "output_type": "stream",
57 | "text": [
58 | "S1\n",
59 | "[[[3 3 3]]]\n",
60 | "[[[3 3 3]]]\n",
61 | "\n",
62 | "S2\n",
63 | "[[[3 3 3]\n",
64 | " [4 4 4]]]\n",
65 | "[[[3 3 3]\n",
66 | " [4 4 4]]]\n",
67 | "\n",
68 | "S3\n",
69 | "[[[3 3 3]]\n",
70 | "\n",
71 | " [[5 5 5]]]\n",
72 | "-------------\n",
73 | "[[[3 3 3]]\n",
74 | "\n",
75 | " [[5 5 5]]]\n"
76 | ]
77 | }
78 | ],
79 | "source": [
80 | "data = np.array([[[1, 1, 1], [2, 2, 2]],\n",
81 | " [[3, 3, 3], [4, 4, 4]],\n",
82 | " [[5, 5, 5], [6, 6, 6]]])\n",
83 | "\n",
84 | "\n",
85 | "s1 = tf.slice(data, [1, 0, 0], [1, 1, 3])\n",
86 | "s2 = tf.slice(data, [1, 0, 0], [1, 2, 3])\n",
87 | "s3 = tf.slice(data, [1, 0, 0], [2, 1, 3])\n",
88 | "\n",
89 | "print('S1')\n",
90 | "print(s1.eval())\n",
91 | "print(data[1:1+1, 0:0+1, 0:0+3])\n",
92 | "\n",
93 | "print('\\nS2')\n",
94 | "print(s2.eval())\n",
95 | "print(data[1:1+1, 0:0+2, 0:0+3])\n",
96 | "\n",
97 | "print('\\nS3')\n",
98 | "print(s3.eval())\n",
99 | "print('-------------')\n",
100 | "print(data[1:1+2, 0:0+1, 0:0+3])"
101 | ]
102 | },
103 | {
104 | "cell_type": "markdown",
105 | "metadata": {},
106 | "source": [
107 | "## tf.split(value, num_or_size_splits, axis=0, num=None, name='split')"
108 | ]
109 | },
110 | {
111 | "cell_type": "code",
112 | "execution_count": 205,
113 | "metadata": {
114 | "collapsed": false
115 | },
116 | "outputs": [
117 | {
118 | "name": "stdout",
119 | "output_type": "stream",
120 | "text": [
121 | "Data\n",
122 | "[[ 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17\n",
123 | " 18 19 20 21 22 23 24 25 26 27 28 29]\n",
124 | " [ 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47\n",
125 | " 48 49 50 51 52 53 54 55 56 57 58 59]\n",
126 | " [ 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77\n",
127 | " 78 79 80 81 82 83 84 85 86 87 88 89]\n",
128 | " [ 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107\n",
129 | " 108 109 110 111 112 113 114 115 116 117 118 119]\n",
130 | " [120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137\n",
131 | " 138 139 140 141 142 143 144 145 146 147 148 149]]\n",
132 | "\n",
133 | "Splitted 01\n",
134 | "[5 4]\n",
135 | "[[ 0 1 2 3]\n",
136 | " [ 30 31 32 33]\n",
137 | " [ 60 61 62 63]\n",
138 | " [ 90 91 92 93]\n",
139 | " [120 121 122 123]]\n",
140 | "\n",
141 | "Splitted 02\n",
142 | "[ 5 15]\n",
143 | "[[ 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18]\n",
144 | " [ 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48]\n",
145 | " [ 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78]\n",
146 | " [ 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108]\n",
147 | " [124 125 126 127 128 129 130 131 132 133 134 135 136 137 138]]\n",
148 | "\n",
149 | "Splitted 03\n",
150 | "[ 5 11]\n",
151 | "[[ 19 20 21 22 23 24 25 26 27 28 29]\n",
152 | " [ 49 50 51 52 53 54 55 56 57 58 59]\n",
153 | " [ 79 80 81 82 83 84 85 86 87 88 89]\n",
154 | " [109 110 111 112 113 114 115 116 117 118 119]\n",
155 | " [139 140 141 142 143 144 145 146 147 148 149]]\n"
156 | ]
157 | }
158 | ],
159 | "source": [
160 | "data = np.arange(5*30).reshape((5, 30))\n",
161 | "print('Data')\n",
162 | "print(data)\n",
163 | "\n",
164 | "splitted1 = tf.split(data, [4, 15, 11], axis=1)\n",
165 | "\n",
166 | "print('\\nSplitted 01')\n",
167 | "print(tf.shape(splitted1[0]).eval()) # 4\n",
168 | "print(splitted1[0].eval())\n",
169 | "\n",
170 | "print('\\nSplitted 02')\n",
171 | "print(tf.shape(splitted1[1]).eval()) # 15\n",
172 | "print(splitted1[1].eval())\n",
173 | "\n",
174 | "print('\\nSplitted 03')\n",
175 | "print(tf.shape(splitted1[2]).eval()) # 11\n",
176 | "print(splitted1[2].eval())"
177 | ]
178 | },
179 | {
180 | "cell_type": "code",
181 | "execution_count": 203,
182 | "metadata": {
183 | "collapsed": false
184 | },
185 | "outputs": [
186 | {
187 | "name": "stdout",
188 | "output_type": "stream",
189 | "text": [
190 | "\n",
191 | "\n",
192 | "Splitted 01\n",
193 | "Shape: [ 1 30]\n",
194 | "[[ 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23\n",
195 | " 24 25 26 27 28 29]]\n",
196 | "numpy\n",
197 | "[[ 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23\n",
198 | " 24 25 26 27 28 29]]\n",
199 | "\n",
200 | "\n",
201 | "Splitted 02\n",
202 | "Shape: [ 2 30]\n",
203 | "[[30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53\n",
204 | " 54 55 56 57 58 59]\n",
205 | " [60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83\n",
206 | " 84 85 86 87 88 89]]\n",
207 | "numpy\n",
208 | "[[30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53\n",
209 | " 54 55 56 57 58 59]\n",
210 | " [60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83\n",
211 | " 84 85 86 87 88 89]]\n",
212 | "\n",
213 | "\n",
214 | "Splitted 03\n",
215 | "Shape: [ 2 30]\n",
216 | "[[ 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107\n",
217 | " 108 109 110 111 112 113 114 115 116 117 118 119]\n",
218 | " [120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137\n",
219 | " 138 139 140 141 142 143 144 145 146 147 148 149]]\n",
220 | "numpy\n",
221 | "[[ 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107\n",
222 | " 108 109 110 111 112 113 114 115 116 117 118 119]\n",
223 | " [120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137\n",
224 | " 138 139 140 141 142 143 144 145 146 147 148 149]]\n"
225 | ]
226 | }
227 | ],
228 | "source": [
229 | "splitted = tf.split(data, [1, 2, 2], axis=0)\n",
230 | "\n",
231 | "print('\\n\\nSplitted 01')\n",
232 | "print('Shape:', tf.shape(splitted[0]).eval())\n",
233 | "print(splitted[0].eval())\n",
234 | "print('numpy')\n",
235 | "print(data[0:1])\n",
236 | "\n",
237 | "print('\\n\\nSplitted 02')\n",
238 | "print('Shape:', tf.shape(splitted[1]).eval())\n",
239 | "print(splitted[1].eval())\n",
240 | "print('numpy')\n",
241 | "print(data[1:3])\n",
242 | "\n",
243 | "print('\\n\\nSplitted 03')\n",
244 | "print('Shape:', tf.shape(splitted[2]).eval())\n",
245 | "print(splitted[2].eval())\n",
246 | "print('numpy')\n",
247 | "print(data[3:5])"
248 | ]
249 | },
250 | {
251 | "cell_type": "code",
252 | "execution_count": null,
253 | "metadata": {
254 | "collapsed": true
255 | },
256 | "outputs": [],
257 | "source": []
258 | }
259 | ],
260 | "metadata": {
261 | "kernelspec": {
262 | "display_name": "Python 3",
263 | "language": "python",
264 | "name": "python3"
265 | },
266 | "language_info": {
267 | "codemirror_mode": {
268 | "name": "ipython",
269 | "version": 3
270 | },
271 | "file_extension": ".py",
272 | "mimetype": "text/x-python",
273 | "name": "python",
274 | "nbconvert_exporter": "python",
275 | "pygments_lexer": "ipython3",
276 | "version": "3.6.0"
277 | }
278 | },
279 | "nbformat": 4,
280 | "nbformat_minor": 2
281 | }
282 |
--------------------------------------------------------------------------------
/202 Logistic Regression with Regularization/titanic-logistic-regression-with-tensorflow.ipynb:
--------------------------------------------------------------------------------
1 | {
2 | "cells": [
3 | {
4 | "cell_type": "code",
5 | "execution_count": 19,
6 | "metadata": {},
7 | "outputs": [
8 | {
9 | "name": "stdout",
10 | "output_type": "stream",
11 | "text": [
12 | "Using matplotlib backend: TkAgg\n",
13 | "Populating the interactive namespace from numpy and matplotlib\n"
14 | ]
15 | }
16 | ],
17 | "source": [
18 | "%pylab \n",
19 | "import numpy as np\n",
20 | "import pandas as pd\n",
21 | "import tensorflow as tf\n",
22 | "\n",
23 | "from sklearn.preprocessing import StandardScaler, MinMaxScaler\n",
24 | "from sklearn.model_selection import train_test_split\n",
25 | "from sklearn.linear_model import LogisticRegression\n",
26 | "from IPython.display import display\n",
27 | "\n",
28 | "np.random.seed(0)"
29 | ]
30 | },
31 | {
32 | "cell_type": "markdown",
33 | "metadata": {},
34 | "source": [
35 | "# Data & Preprocessing\n",
36 | "\n",
37 | "\n",
38 | "| Variable | Definition | Key |\n",
39 | "|:---------|:-----------|:----|\n",
40 | "| survival | 생존 | 0 = No, 1 = Yes | \n",
41 | "| pclass | 티켓 등급 | 1 = 1st, 2 = 2nd, 3 = 3rd | \n",
42 | "| sex | 성별 | | \n",
43 | "| Age | 나이 | | \n",
44 | "| sibsp\t | 타이타닉에 함께 타고 있는 형제, 자매 또는 배우 (갯수) | | \n",
45 | "| parch | 타이타닉에 함께 타고 있는 부모 또는 자식 (갯수) | | \n",
46 | "| ticket | 티켓 넘버 | | \n",
47 | "| fare | 티켓 가격\t | | \n",
48 | "| cabin | 객실 번호 | | \n",
49 | "| embarked | 출항지 | C = Cherbourg, Q = Queenstown, S = Southampton | "
50 | ]
51 | },
52 | {
53 | "cell_type": "code",
54 | "execution_count": 2,
55 | "metadata": {},
56 | "outputs": [
57 | {
58 | "data": {
59 | "text/html": [
60 | "\n",
61 | "\n",
74 | "
\n",
75 | " \n",
76 | " \n",
77 | " | \n",
78 | " Survived | \n",
79 | " Age | \n",
80 | " SibSp | \n",
81 | " Parch | \n",
82 | " Fare | \n",
83 | " Pclass_1 | \n",
84 | " Pclass_2 | \n",
85 | " Pclass_3 | \n",
86 | " Sex_female | \n",
87 | " Sex_male | \n",
88 | " Embarked_C | \n",
89 | " Embarked_Q | \n",
90 | " Embarked_S | \n",
91 | "
\n",
92 | " \n",
93 | " \n",
94 | " \n",
95 | " | 0 | \n",
96 | " 0 | \n",
97 | " 22.0 | \n",
98 | " 1 | \n",
99 | " 0 | \n",
100 | " 7.2500 | \n",
101 | " 0 | \n",
102 | " 0 | \n",
103 | " 1 | \n",
104 | " 0 | \n",
105 | " 1 | \n",
106 | " 0 | \n",
107 | " 0 | \n",
108 | " 1 | \n",
109 | "
\n",
110 | " \n",
111 | " | 1 | \n",
112 | " 1 | \n",
113 | " 38.0 | \n",
114 | " 1 | \n",
115 | " 0 | \n",
116 | " 71.2833 | \n",
117 | " 1 | \n",
118 | " 0 | \n",
119 | " 0 | \n",
120 | " 1 | \n",
121 | " 0 | \n",
122 | " 1 | \n",
123 | " 0 | \n",
124 | " 0 | \n",
125 | "
\n",
126 | " \n",
127 | " | 2 | \n",
128 | " 1 | \n",
129 | " 26.0 | \n",
130 | " 0 | \n",
131 | " 0 | \n",
132 | " 7.9250 | \n",
133 | " 0 | \n",
134 | " 0 | \n",
135 | " 1 | \n",
136 | " 1 | \n",
137 | " 0 | \n",
138 | " 0 | \n",
139 | " 0 | \n",
140 | " 1 | \n",
141 | "
\n",
142 | " \n",
143 | " | 3 | \n",
144 | " 1 | \n",
145 | " 35.0 | \n",
146 | " 1 | \n",
147 | " 0 | \n",
148 | " 53.1000 | \n",
149 | " 1 | \n",
150 | " 0 | \n",
151 | " 0 | \n",
152 | " 1 | \n",
153 | " 0 | \n",
154 | " 0 | \n",
155 | " 0 | \n",
156 | " 1 | \n",
157 | "
\n",
158 | " \n",
159 | " | 4 | \n",
160 | " 0 | \n",
161 | " 35.0 | \n",
162 | " 0 | \n",
163 | " 0 | \n",
164 | " 8.0500 | \n",
165 | " 0 | \n",
166 | " 0 | \n",
167 | " 1 | \n",
168 | " 0 | \n",
169 | " 1 | \n",
170 | " 0 | \n",
171 | " 0 | \n",
172 | " 1 | \n",
173 | "
\n",
174 | " \n",
175 | "
\n",
176 | "
\n",
255 | "\n",
268 | "
\n",
269 | " \n",
270 | " \n",
271 | " | \n",
272 | " Survived | \n",
273 | " Age | \n",
274 | " SibSp | \n",
275 | " Parch | \n",
276 | " Fare | \n",
277 | " Pclass_1 | \n",
278 | " Pclass_2 | \n",
279 | " Pclass_3 | \n",
280 | " Sex_female | \n",
281 | " Sex_male | \n",
282 | " Embarked_C | \n",
283 | " Embarked_Q | \n",
284 | " Embarked_S | \n",
285 | "
\n",
286 | " \n",
287 | " \n",
288 | " \n",
289 | "
\n",
290 | "
\n",
324 | "\n",
337 | "
\n",
338 | " \n",
339 | " \n",
340 | " | \n",
341 | " Survived | \n",
342 | " Age | \n",
343 | " SibSp | \n",
344 | " Parch | \n",
345 | " Fare | \n",
346 | " Pclass_1 | \n",
347 | " Pclass_2 | \n",
348 | " Pclass_3 | \n",
349 | " Sex_female | \n",
350 | " Sex_male | \n",
351 | " Embarked_C | \n",
352 | " Embarked_Q | \n",
353 | " Embarked_S | \n",
354 | "
\n",
355 | " \n",
356 | " \n",
357 | " \n",
358 | " | 0 | \n",
359 | " 0 | \n",
360 | " 0.2750 | \n",
361 | " 0.125 | \n",
362 | " 0.0 | \n",
363 | " 0.014151 | \n",
364 | " 0 | \n",
365 | " 0 | \n",
366 | " 1 | \n",
367 | " 0 | \n",
368 | " 1 | \n",
369 | " 0 | \n",
370 | " 0 | \n",
371 | " 1 | \n",
372 | "
\n",
373 | " \n",
374 | " | 1 | \n",
375 | " 1 | \n",
376 | " 0.4750 | \n",
377 | " 0.125 | \n",
378 | " 0.0 | \n",
379 | " 0.139136 | \n",
380 | " 1 | \n",
381 | " 0 | \n",
382 | " 0 | \n",
383 | " 1 | \n",
384 | " 0 | \n",
385 | " 1 | \n",
386 | " 0 | \n",
387 | " 0 | \n",
388 | "
\n",
389 | " \n",
390 | " | 2 | \n",
391 | " 1 | \n",
392 | " 0.3250 | \n",
393 | " 0.000 | \n",
394 | " 0.0 | \n",
395 | " 0.015469 | \n",
396 | " 0 | \n",
397 | " 0 | \n",
398 | " 1 | \n",
399 | " 1 | \n",
400 | " 0 | \n",
401 | " 0 | \n",
402 | " 0 | \n",
403 | " 1 | \n",
404 | "
\n",
405 | " \n",
406 | " | 3 | \n",
407 | " 1 | \n",
408 | " 0.4375 | \n",
409 | " 0.125 | \n",
410 | " 0.0 | \n",
411 | " 0.103644 | \n",
412 | " 1 | \n",
413 | " 0 | \n",
414 | " 0 | \n",
415 | " 1 | \n",
416 | " 0 | \n",
417 | " 0 | \n",
418 | " 0 | \n",
419 | " 1 | \n",
420 | "
\n",
421 | " \n",
422 | " | 4 | \n",
423 | " 0 | \n",
424 | " 0.4375 | \n",
425 | " 0.000 | \n",
426 | " 0.0 | \n",
427 | " 0.015713 | \n",
428 | " 0 | \n",
429 | " 0 | \n",
430 | " 1 | \n",
431 | " 0 | \n",
432 | " 1 | \n",
433 | " 0 | \n",
434 | " 0 | \n",
435 | " 1 | \n",
436 | "
\n",
437 | " \n",
438 | "
\n",
439 | "
\n",
218 | "2. damping -- save damping "
219 | ]
220 | },
221 | {
222 | "cell_type": "code",
223 | "execution_count": 11,
224 | "metadata": {
225 | "collapsed": false,
226 | "deletable": true,
227 | "editable": true,
228 | "scrolled": false
229 | },
230 | "outputs": [],
231 | "source": [
232 | "eps = tf.placeholder(tf.float32, shape=())\n",
233 | "damping = tf.placeholder(tf.float32, shape=())\n",
234 | "\n",
235 | "# simulation state\n",
236 | "U = tf.Variable(u_init)\n",
237 | "Ut = tf.Variable(ut_init)\n",
238 | "\n",
239 | "# Discretized PDE update rules\n",
240 | "U_ = U + eps * Ut\n",
241 | "Ut_ = Ut + eps * (laplace(U) - damping * Ut)\n",
242 | "\n",
243 | "# Operation to update the state\n",
244 | "step = tf.group(U.assign(U_), Ut.assign(Ut_))"
245 | ]
246 | },
247 | {
248 | "cell_type": "markdown",
249 | "metadata": {
250 | "deletable": true,
251 | "editable": true
252 | },
253 | "source": [
254 | "### Run the simulation"
255 | ]
256 | },
257 | {
258 | "cell_type": "code",
259 | "execution_count": 24,
260 | "metadata": {
261 | "collapsed": false,
262 | "deletable": true,
263 | "editable": true
264 | },
265 | "outputs": [
266 | {
267 | "data": {
268 | "image/jpeg": "/9j/4AAQSkZJRgABAQAAAQABAAD/2wBDAAgGBgcGBQgHBwcJCQgKDBQNDAsLDBkSEw8UHRofHh0a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\n",
269 | "text/plain": [
270 | ""
271 | ]
272 | },
273 | "metadata": {},
274 | "output_type": "display_data"
275 | }
276 | ],
277 | "source": [
278 | "tf.global_variables_initializer().run()\n",
279 | "\n",
280 | "for i in range(1000):\n",
281 | " step.run({eps: 0.03, damping: 0.04})\n",
282 | " display_array(U.eval(), rng=[-0.1, 0.1])"
283 | ]
284 | }
285 | ],
286 | "metadata": {
287 | "kernelspec": {
288 | "display_name": "Python 3",
289 | "language": "python",
290 | "name": "python3"
291 | },
292 | "language_info": {
293 | "codemirror_mode": {
294 | "name": "ipython",
295 | "version": 3
296 | },
297 | "file_extension": ".py",
298 | "mimetype": "text/x-python",
299 | "name": "python",
300 | "nbconvert_exporter": "python",
301 | "pygments_lexer": "ipython3",
302 | "version": "3.6.0"
303 | }
304 | },
305 | "nbformat": 4,
306 | "nbformat_minor": 2
307 | }
308 |
--------------------------------------------------------------------------------