1+ {
2+ "cells" : [
3+ {
4+ "cell_type" : " code"
5+ "execution_count" : null ,
6+ "metadata" : {},
7+ "outputs" : [],
8+ "source" : [
9+ " using Tensorflow;"
10+ ]
11+ },
12+ {
13+ "cell_type" : " code"
14+ "execution_count" : null ,
15+ "metadata" : {},
16+ "outputs" : [],
17+ "source" : [
18+ " using static Tensorflow.Python;"
19+ ]
20+ },
21+ {
22+ "cell_type" : " code"
23+ "execution_count" : null ,
24+ "metadata" : {},
25+ "outputs" : [],
26+ "source" : [
27+ " using PlotNET;"
28+ ]
29+ },
30+ {
31+ "cell_type" : " code"
32+ "execution_count" : null ,
33+ "metadata" : {},
34+ "outputs" : [],
35+ "source" : [
36+ " using NumSharp;"
37+ ]
38+ },
39+ {
40+ "cell_type" : " code"
41+ "execution_count" : null ,
42+ "metadata" : {},
43+ "outputs" : [],
44+ "source" : [
45+ " int training_epochs = 1000;\n "
46+ " \n "
47+ " // Parameters\n "
48+ " float learning_rate = 0.01f;\n "
49+ " int display_step = 50;\n "
50+ " \n "
51+ " NumPyRandom rng = np.random;\n "
52+ " NDArray train_X, train_Y;\n "
53+ " int n_samples;"
54+ ]
55+ },
56+ {
57+ "cell_type" : " code"
58+ "execution_count" : null ,
59+ "metadata" : {},
60+ "outputs" : [],
61+ "source" : [
62+ " train_X = np.array(3.3f, 4.4f, 5.5f, 6.71f, 6.93f, 4.168f, 9.779f, 6.182f, 7.59f, 2.167f,\n "
63+ " 7.042f, 10.791f, 5.313f, 7.997f, 5.654f, 9.27f, 3.1f);\n "
64+ " train_Y = np.array(1.7f, 2.76f, 2.09f, 3.19f, 1.694f, 1.573f, 3.366f, 2.596f, 2.53f, 1.221f,\n "
65+ " 2.827f, 3.465f, 1.65f, 2.904f, 2.42f, 2.94f, 1.3f);\n "
66+ " n_samples = train_X.shape[0];\n "
67+ " \n "
68+ " // tf Graph Input\n "
69+ " var X = tf.placeholder(tf.float32);\n "
70+ " var Y = tf.placeholder(tf.float32);\n "
71+ " \n "
72+ " // Set model weights \n "
73+ " // We can set a fixed init value in order to debug\n "
74+ " // var rnd1 = rng.randn<float>();\n "
75+ " // var rnd2 = rng.randn<float>();\n "
76+ " var W = tf.Variable(-0.06f, name: \" weight\" );\n "
77+ " var b = tf.Variable(-0.73f, name: \" bias\" );\n "
78+ " \n "
79+ " // Construct a linear model\n "
80+ " var pred = tf.add(tf.multiply(X, W), b);\n "
81+ " \n "
82+ " // Mean squared error\n "
83+ " var cost = tf.reduce_sum(tf.pow(pred - Y, 2.0f)) / (2.0f * n_samples);\n "
84+ " \n "
85+ " // Gradient descent\n "
86+ " // Note, minimize() knows to modify W and b because Variable objects are trainable=True by default\n "
87+ " var optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(cost);\n "
88+ " \n "
89+ " // Initialize the variables (i.e. assign their default value)\n "
90+ " var init = tf.global_variables_initializer();\n "
91+ ]
92+ },
93+ {
94+ "cell_type" : " code"
95+ "execution_count" : null ,
96+ "metadata" : {
97+ "scrolled" : false
98+ },
99+ "outputs" : [],
100+ "source" : [
101+ " // Start training\n "
102+ " using (var sess = tf.Session())\n "
103+ " {\n "
104+ " // Run the initializer\n "
105+ " sess.run(init);\n "
106+ " \n "
107+ " // Fit all training data\n "
108+ " for (int epoch = 0; epoch < training_epochs; epoch++)\n "
109+ " {\n "
110+ " foreach (var (x, y) in zip<float>(train_X, train_Y))\n "
111+ " {\n "
112+ " sess.run(optimizer, \n "
113+ " new FeedItem(X, x),\n "
114+ " new FeedItem(Y, y));\n "
115+ " }\n "
116+ " \n "
117+ " // Display logs per epoch step\n "
118+ " if ((epoch + 1) % display_step == 0)\n "
119+ " {\n "
120+ " var c = sess.run(cost, \n "
121+ " new FeedItem(X, train_X),\n "
122+ " new FeedItem(Y, train_Y));\n "
123+ " Console.WriteLine($\" Epoch: {epoch + 1} cost={c} \" + $\" W={sess.run(W)} b={sess.run(b)}\" );\n "
124+ " }\n "
125+ " }\n "
126+ " \n "
127+ " Console.WriteLine(\" Optimization Finished!\" );\n "
128+ " \n "
129+ " var training_cost = sess.run(cost,\n "
130+ " new FeedItem(X, train_X),\n "
131+ " new FeedItem(Y, train_Y));\n "
132+ " \n "
133+ " var plotter = new Plotter();\n "
134+ " \n "
135+ " plotter.Plot(\n "
136+ " train_X,\n "
137+ " train_Y,\n "
138+ " \" Original data\" , ChartType.Scatter,\" markers\" );\n "
139+ " plotter.Plot(\n "
140+ " train_X,\n "
141+ " sess.run(W) * train_X + sess.run(b),\n "
142+ " \" Fitted line\" , ChartType.Scatter, \" Fitted line\" );\n "
143+ " \n "
144+ " plotter.Show();\n "
145+ " \n "
146+ " // Testing example\n "
147+ " var test_X = np.array(6.83f, 4.668f, 8.9f, 7.91f, 5.7f, 8.7f, 3.1f, 2.1f);\n "
148+ " var test_Y = np.array(1.84f, 2.273f, 3.2f, 2.831f, 2.92f, 3.24f, 1.35f, 1.03f);\n "
149+ " \n "
150+ " Console.WriteLine(\" Testing... (Mean square loss Comparison)\" );\n "
151+ " \n "
152+ " var testing_cost = sess.run(tf.reduce_sum(tf.pow(pred - Y, 2.0f)) / (2.0f * test_X.shape[0]),\n "
153+ " new FeedItem(X, test_X), \n "
154+ " new FeedItem(Y, test_Y));\n "
155+ " \n "
156+ " Console.WriteLine($\" Testing cost={testing_cost}\" );\n "
157+ " \n "
158+ " var diff = Math.Abs((float)training_cost - (float)testing_cost);\n "
159+ " Console.WriteLine($\" Absolute mean square loss difference: {diff}\" );\n "
160+ " \n "
161+ " plotter.Plot(\n "
162+ " test_X,\n "
163+ " test_Y,\n "
164+ " \" Testing data\" , ChartType.Scatter, \" markers\" );\n "
165+ " plotter.Plot(\n "
166+ " train_X,\n "
167+ " sess.run(W) * train_X + sess.run(b),\n "
168+ " \" Fitted line\" , ChartType.Scatter);\n "
169+ " \n "
170+ " plotter.Show();\n "
171+ " \n "
172+ " return diff < 0.01;\n "
173+ " }"
174+ ]
175+ },
176+ {
177+ "cell_type" : " code"
178+ "execution_count" : null ,
179+ "metadata" : {},
180+ "outputs" : [],
181+ "source" : []
182+ }
183+ ],
184+ "metadata" : {
185+ "kernelspec" : {
186+ "display_name" : " SciSharp Cube"
187+ "language" : " csharp"
188+ "name" : " csharpcore"
189+ },
190+ "language_info" : {
191+ "file_extension" : " .cs"
192+ "mimetype" : " text/x-csharp"
193+ "name" : " .netstandard"
194+ "pygments_lexer" : " CSharp"
195+ "version" : " 4.0.30319"
196+ }
197+ },
198+ "nbformat" : 4 ,
199+ "nbformat_minor" : 2
200+ }
0 commit comments