質問編集履歴
2
コードブロックの使用
test
CHANGED
File without changes
|
test
CHANGED
@@ -2,316 +2,318 @@
|
|
2
2
|
|
3
3
|
|
4
4
|
|
5
|
+
|
6
|
+
|
7
|
+
pythonで機械学習のプログラムを作るうえで
|
8
|
+
|
9
|
+
一変数ではなく、多変数のデータを読み込んで実行したい
|
10
|
+
|
11
|
+
|
12
|
+
|
13
|
+
一変数でのプログラムは作って実行もできたので、そこに何かを付け足せばいいのか
|
14
|
+
|
15
|
+
または新しく作り直さなければならないのかを知りたく、またそうするにはどうすればいいかを教えてほしいです
|
16
|
+
|
17
|
+
|
18
|
+
|
19
|
+
|
20
|
+
|
21
|
+
### 該当のソースコード
|
22
|
+
|
23
|
+
|
24
|
+
|
25
|
+
|
26
|
+
|
27
|
+
|
28
|
+
|
29
|
+
```python
|
30
|
+
|
31
|
+
import tensorflow as tf
|
32
|
+
|
33
|
+
import numpy as np
|
34
|
+
|
35
|
+
import random
|
36
|
+
|
37
|
+
|
38
|
+
|
39
|
+
|
40
|
+
|
41
|
+
num_of_input_nodes = 1
|
42
|
+
|
43
|
+
num_of_hidden_nodes = 80
|
44
|
+
|
45
|
+
num_of_output_nodes = 1
|
46
|
+
|
47
|
+
length_of_sequences = 10 ← 学習・識別時の時系列データ数(area.txtとarea2.txtの列数)
|
48
|
+
|
49
|
+
num_of_training_epochs = 5000 ← 学習回数
|
50
|
+
|
51
|
+
size_of_mini_batch = 10
|
52
|
+
|
53
|
+
num_of_prediction_epochs = 107 ← 識別ファイル(area2.txtとseikai2.txtの行数)
|
54
|
+
|
55
|
+
learning_rate = 0.01
|
56
|
+
|
57
|
+
forget_bias = 0.8
|
58
|
+
|
59
|
+
num_of_sample = 109 ← 学習ファイル(area.txtとseikai.txtの行数)
|
60
|
+
|
61
|
+
|
62
|
+
|
63
|
+
|
64
|
+
|
65
|
+
def get_batch(batch_size, X, t):
|
66
|
+
|
67
|
+
rnum = [random.randint(0, len(X) - 1) for x in range(batch_size)]
|
68
|
+
|
69
|
+
xs = np.array([[[y] for y in list(X[r])] for r in rnum])
|
70
|
+
|
71
|
+
ts = np.array([[t[r]] for r in rnum])
|
72
|
+
|
73
|
+
return xs, ts
|
74
|
+
|
75
|
+
|
76
|
+
|
77
|
+
|
78
|
+
|
79
|
+
def create_data(nb_of_samples, sequence_len):
|
80
|
+
|
81
|
+
X = np.loadtxt('area.txt') ← 学習データファイルの読込み
|
82
|
+
|
83
|
+
t = np.loadtxt('seikai.txt') ← 学習正解値ファイルの読込み
|
84
|
+
|
85
|
+
print(X)
|
86
|
+
|
87
|
+
print(t)
|
88
|
+
|
89
|
+
return X, t
|
90
|
+
|
91
|
+
|
92
|
+
|
93
|
+
|
94
|
+
|
95
|
+
def make_prediction(nb_of_samples):
|
96
|
+
|
97
|
+
xs = np.loadtxt('area2.txt') ← 識別データファイルの読込み
|
98
|
+
|
99
|
+
ts = np.loadtxt('seikai2.txt') ← 識別正解値データファイルの読込み(正解率計算用。実際の使用時は正解値は使用しない)
|
100
|
+
|
101
|
+
return np.array([[[y] for y in x] for x in xs]), np.array([[x] for x in ts])
|
102
|
+
|
103
|
+
|
104
|
+
|
105
|
+
|
106
|
+
|
107
|
+
def inference(input_ph, istate_ph):
|
108
|
+
|
109
|
+
with tf.name_scope("inference") as scope:
|
110
|
+
|
111
|
+
weight1_var = tf.Variable(tf.truncated_normal(
|
112
|
+
|
113
|
+
[num_of_input_nodes, num_of_hidden_nodes], stddev=0.1), name="weight1")
|
114
|
+
|
115
|
+
weight2_var = tf.Variable(tf.truncated_normal(
|
116
|
+
|
117
|
+
[num_of_hidden_nodes, num_of_output_nodes], stddev=0.1), name="weight2")
|
118
|
+
|
119
|
+
bias1_var = tf.Variable(tf.truncated_normal([num_of_hidden_nodes], stddev=0.1), name="bias1")
|
120
|
+
|
121
|
+
bias2_var = tf.Variable(tf.truncated_normal([num_of_output_nodes], stddev=0.1), name="bias2")
|
122
|
+
|
123
|
+
|
124
|
+
|
125
|
+
in1 = tf.transpose(input_ph, [1, 0, 2])
|
126
|
+
|
127
|
+
in2 = tf.reshape(in1, [-1, num_of_input_nodes])
|
128
|
+
|
129
|
+
in3 = tf.matmul(in2, weight1_var) + bias1_var
|
130
|
+
|
131
|
+
in4 = tf.split(in3, length_of_sequences, 0)
|
132
|
+
|
133
|
+
|
134
|
+
|
135
|
+
cell = tf.nn.rnn_cell.BasicLSTMCell(num_of_hidden_nodes, forget_bias=forget_bias, state_is_tuple=False)
|
136
|
+
|
137
|
+
rnn_output, states_op = tf.contrib.rnn.static_rnn(cell, in4, initial_state=istate_ph)
|
138
|
+
|
139
|
+
output_op = tf.matmul(rnn_output[-1], weight2_var) + bias2_var
|
140
|
+
|
141
|
+
|
142
|
+
|
143
|
+
# Add summary ops to collect data
|
144
|
+
|
145
|
+
w1_hist = tf.summary.histogram("weights1", weight1_var)
|
146
|
+
|
147
|
+
w2_hist = tf.summary.histogram("weights2", weight2_var)
|
148
|
+
|
149
|
+
b1_hist = tf.summary.histogram("biases1", bias1_var)
|
150
|
+
|
151
|
+
b2_hist = tf.summary.histogram("biases2", bias2_var)
|
152
|
+
|
153
|
+
output_hist = tf.summary.histogram("output", output_op)
|
154
|
+
|
155
|
+
results = [weight1_var, weight2_var, bias1_var, bias2_var]
|
156
|
+
|
157
|
+
return output_op, states_op, results
|
158
|
+
|
159
|
+
|
160
|
+
|
161
|
+
|
162
|
+
|
163
|
+
def loss(output_op, supervisor_ph):
|
164
|
+
|
165
|
+
with tf.name_scope("loss") as scope:
|
166
|
+
|
167
|
+
square_error = tf.reduce_mean(tf.square(output_op - supervisor_ph))
|
168
|
+
|
169
|
+
loss_op = square_error
|
170
|
+
|
171
|
+
tf.summary.scalar("loss", loss_op)
|
172
|
+
|
173
|
+
return loss_op
|
174
|
+
|
175
|
+
|
176
|
+
|
177
|
+
|
178
|
+
|
179
|
+
def training(loss_op):
|
180
|
+
|
181
|
+
with tf.name_scope("training") as scope:
|
182
|
+
|
183
|
+
training_op = optimizer.minimize(loss_op)
|
184
|
+
|
185
|
+
return training_op
|
186
|
+
|
187
|
+
|
188
|
+
|
189
|
+
|
190
|
+
|
191
|
+
def calc_accuracy(output_op, prints=False):
|
192
|
+
|
193
|
+
inputs, ts = make_prediction(num_of_prediction_epochs)
|
194
|
+
|
195
|
+
pred_dict = {
|
196
|
+
|
197
|
+
input_ph: inputs,
|
198
|
+
|
199
|
+
supervisor_ph: ts,
|
200
|
+
|
201
|
+
istate_ph: np.zeros((num_of_prediction_epochs, num_of_hidden_nodes * 2)),
|
202
|
+
|
203
|
+
}
|
204
|
+
|
205
|
+
output = sess.run([output_op], feed_dict=pred_dict)
|
206
|
+
|
207
|
+
|
208
|
+
|
209
|
+
def print_result(i, p, q):
|
210
|
+
|
211
|
+
[print(list(x)[0]) for x in i]
|
212
|
+
|
213
|
+
print("output: %f, correct: %d" % (p, q))
|
214
|
+
|
215
|
+
if prints:
|
216
|
+
|
217
|
+
[print_result(i, p, q) for i, p, q in zip(inputs, output[0], ts)]
|
218
|
+
|
219
|
+
|
220
|
+
|
221
|
+
opt = abs(output - ts)[0]
|
222
|
+
|
223
|
+
total = sum([1 if x[0] < 0.5 else 0 for x in opt]) ← 0.05から0.5に変更
|
224
|
+
|
225
|
+
print("accuracy %f" % (total / float(len(ts))))
|
226
|
+
|
227
|
+
return output
|
228
|
+
|
229
|
+
|
230
|
+
|
231
|
+
random.seed(0)
|
232
|
+
|
233
|
+
np.random.seed(0)
|
234
|
+
|
235
|
+
tf.set_random_seed(0)
|
236
|
+
|
237
|
+
|
238
|
+
|
239
|
+
optimizer = tf.train.GradientDescentOptimizer(learning_rate=learning_rate)
|
240
|
+
|
241
|
+
|
242
|
+
|
243
|
+
X, t = create_data(num_of_sample, length_of_sequences)
|
244
|
+
|
245
|
+
|
246
|
+
|
247
|
+
with tf.Graph().as_default():
|
248
|
+
|
249
|
+
input_ph = tf.placeholder(tf.float32, [None, length_of_sequences, num_of_input_nodes], name="input")
|
250
|
+
|
251
|
+
supervisor_ph = tf.placeholder(tf.float32, [None, num_of_output_nodes], name="supervisor")
|
252
|
+
|
253
|
+
istate_ph = tf.placeholder(tf.float32, [None, num_of_hidden_nodes * 2], name="istate")
|
254
|
+
|
255
|
+
|
256
|
+
|
257
|
+
output_op, states_op, datas_op = inference(input_ph, istate_ph)
|
258
|
+
|
259
|
+
loss_op = loss(output_op, supervisor_ph)
|
260
|
+
|
261
|
+
training_op = training(loss_op)
|
262
|
+
|
263
|
+
|
264
|
+
|
265
|
+
summary_op = tf.summary.merge_all()
|
266
|
+
|
267
|
+
init = tf.initialize_all_variables()
|
268
|
+
|
269
|
+
|
270
|
+
|
271
|
+
with tf.Session() as sess:
|
272
|
+
|
273
|
+
saver = tf.train.Saver()
|
274
|
+
|
275
|
+
summary_writer = tf.summary.FileWriter("/tmp/tensorflow_log", graph=sess.graph)
|
276
|
+
|
277
|
+
sess.run(init)
|
278
|
+
|
279
|
+
|
280
|
+
|
281
|
+
for epoch in range(num_of_training_epochs):
|
282
|
+
|
283
|
+
inputs, supervisors = get_batch(size_of_mini_batch, X, t)
|
284
|
+
|
285
|
+
train_dict = {
|
286
|
+
|
287
|
+
input_ph: inputs,
|
288
|
+
|
289
|
+
supervisor_ph: supervisors,
|
290
|
+
|
291
|
+
istate_ph: np.zeros((size_of_mini_batch, num_of_hidden_nodes * 2)),
|
292
|
+
|
293
|
+
}
|
294
|
+
|
295
|
+
sess.run(training_op, feed_dict=train_dict)
|
296
|
+
|
297
|
+
|
298
|
+
|
299
|
+
if (epoch) % 100 == 0:
|
300
|
+
|
301
|
+
summary_str, train_loss = sess.run([summary_op, loss_op], feed_dict=train_dict)
|
302
|
+
|
303
|
+
print("train#%d, train loss: %e" % (epoch, train_loss))
|
304
|
+
|
305
|
+
summary_writer.add_summary(summary_str, epoch)
|
306
|
+
|
307
|
+
if (epoch) % 500 == 0:
|
308
|
+
|
309
|
+
calc_accuracy(output_op)
|
310
|
+
|
311
|
+
|
312
|
+
|
313
|
+
calc_accuracy(output_op, prints=True)
|
314
|
+
|
315
|
+
datas = sess.run(datas_op)
|
316
|
+
|
317
|
+
saver.save(sess, "model.ckpt")
|
318
|
+
|
5
319
|
```
|
6
|
-
|
7
|
-
python
|
8
|
-
|
9
|
-
```で機械学習のプログラムを作るうえで
|
10
|
-
|
11
|
-
一変数ではなく、多変数のデータを読み込んで実行したい
|
12
|
-
|
13
|
-
|
14
|
-
|
15
|
-
一変数でのプログラムは作って実行もできたので、そこに何かを付け足せばいいのか
|
16
|
-
|
17
|
-
または新しく作り直さなければならないのかを知りたく、またそうするにはどうすればいいかを教えてほしいです
|
18
|
-
|
19
|
-
|
20
|
-
|
21
|
-
|
22
|
-
|
23
|
-
### 該当のソースコード
|
24
|
-
|
25
|
-
|
26
|
-
|
27
|
-
|
28
|
-
|
29
|
-
|
30
|
-
|
31
|
-
import tensorflow as tf
|
32
|
-
|
33
|
-
import numpy as np
|
34
|
-
|
35
|
-
import random
|
36
|
-
|
37
|
-
|
38
|
-
|
39
|
-
|
40
|
-
|
41
|
-
num_of_input_nodes = 1
|
42
|
-
|
43
|
-
num_of_hidden_nodes = 80
|
44
|
-
|
45
|
-
num_of_output_nodes = 1
|
46
|
-
|
47
|
-
length_of_sequences = 10 ← 学習・識別時の時系列データ数(area.txtとarea2.txtの列数)
|
48
|
-
|
49
|
-
num_of_training_epochs = 5000 ← 学習回数
|
50
|
-
|
51
|
-
size_of_mini_batch = 10
|
52
|
-
|
53
|
-
num_of_prediction_epochs = 107 ← 識別ファイル(area2.txtとseikai2.txtの行数)
|
54
|
-
|
55
|
-
learning_rate = 0.01
|
56
|
-
|
57
|
-
forget_bias = 0.8
|
58
|
-
|
59
|
-
num_of_sample = 109 ← 学習ファイル(area.txtとseikai.txtの行数)
|
60
|
-
|
61
|
-
|
62
|
-
|
63
|
-
|
64
|
-
|
65
|
-
def get_batch(batch_size, X, t):
|
66
|
-
|
67
|
-
rnum = [random.randint(0, len(X) - 1) for x in range(batch_size)]
|
68
|
-
|
69
|
-
xs = np.array([[[y] for y in list(X[r])] for r in rnum])
|
70
|
-
|
71
|
-
ts = np.array([[t[r]] for r in rnum])
|
72
|
-
|
73
|
-
return xs, ts
|
74
|
-
|
75
|
-
|
76
|
-
|
77
|
-
|
78
|
-
|
79
|
-
def create_data(nb_of_samples, sequence_len):
|
80
|
-
|
81
|
-
X = np.loadtxt('area.txt') ← 学習データファイルの読込み
|
82
|
-
|
83
|
-
t = np.loadtxt('seikai.txt') ← 学習正解値ファイルの読込み
|
84
|
-
|
85
|
-
print(X)
|
86
|
-
|
87
|
-
print(t)
|
88
|
-
|
89
|
-
return X, t
|
90
|
-
|
91
|
-
|
92
|
-
|
93
|
-
|
94
|
-
|
95
|
-
def make_prediction(nb_of_samples):
|
96
|
-
|
97
|
-
xs = np.loadtxt('area2.txt') ← 識別データファイルの読込み
|
98
|
-
|
99
|
-
ts = np.loadtxt('seikai2.txt') ← 識別正解値データファイルの読込み(正解率計算用。実際の使用時は正解値は使用しない)
|
100
|
-
|
101
|
-
return np.array([[[y] for y in x] for x in xs]), np.array([[x] for x in ts])
|
102
|
-
|
103
|
-
|
104
|
-
|
105
|
-
|
106
|
-
|
107
|
-
def inference(input_ph, istate_ph):
|
108
|
-
|
109
|
-
with tf.name_scope("inference") as scope:
|
110
|
-
|
111
|
-
weight1_var = tf.Variable(tf.truncated_normal(
|
112
|
-
|
113
|
-
[num_of_input_nodes, num_of_hidden_nodes], stddev=0.1), name="weight1")
|
114
|
-
|
115
|
-
weight2_var = tf.Variable(tf.truncated_normal(
|
116
|
-
|
117
|
-
[num_of_hidden_nodes, num_of_output_nodes], stddev=0.1), name="weight2")
|
118
|
-
|
119
|
-
bias1_var = tf.Variable(tf.truncated_normal([num_of_hidden_nodes], stddev=0.1), name="bias1")
|
120
|
-
|
121
|
-
bias2_var = tf.Variable(tf.truncated_normal([num_of_output_nodes], stddev=0.1), name="bias2")
|
122
|
-
|
123
|
-
|
124
|
-
|
125
|
-
in1 = tf.transpose(input_ph, [1, 0, 2])
|
126
|
-
|
127
|
-
in2 = tf.reshape(in1, [-1, num_of_input_nodes])
|
128
|
-
|
129
|
-
in3 = tf.matmul(in2, weight1_var) + bias1_var
|
130
|
-
|
131
|
-
in4 = tf.split(in3, length_of_sequences, 0)
|
132
|
-
|
133
|
-
|
134
|
-
|
135
|
-
cell = tf.nn.rnn_cell.BasicLSTMCell(num_of_hidden_nodes, forget_bias=forget_bias, state_is_tuple=False)
|
136
|
-
|
137
|
-
rnn_output, states_op = tf.contrib.rnn.static_rnn(cell, in4, initial_state=istate_ph)
|
138
|
-
|
139
|
-
output_op = tf.matmul(rnn_output[-1], weight2_var) + bias2_var
|
140
|
-
|
141
|
-
|
142
|
-
|
143
|
-
# Add summary ops to collect data
|
144
|
-
|
145
|
-
w1_hist = tf.summary.histogram("weights1", weight1_var)
|
146
|
-
|
147
|
-
w2_hist = tf.summary.histogram("weights2", weight2_var)
|
148
|
-
|
149
|
-
b1_hist = tf.summary.histogram("biases1", bias1_var)
|
150
|
-
|
151
|
-
b2_hist = tf.summary.histogram("biases2", bias2_var)
|
152
|
-
|
153
|
-
output_hist = tf.summary.histogram("output", output_op)
|
154
|
-
|
155
|
-
results = [weight1_var, weight2_var, bias1_var, bias2_var]
|
156
|
-
|
157
|
-
return output_op, states_op, results
|
158
|
-
|
159
|
-
|
160
|
-
|
161
|
-
|
162
|
-
|
163
|
-
def loss(output_op, supervisor_ph):
|
164
|
-
|
165
|
-
with tf.name_scope("loss") as scope:
|
166
|
-
|
167
|
-
square_error = tf.reduce_mean(tf.square(output_op - supervisor_ph))
|
168
|
-
|
169
|
-
loss_op = square_error
|
170
|
-
|
171
|
-
tf.summary.scalar("loss", loss_op)
|
172
|
-
|
173
|
-
return loss_op
|
174
|
-
|
175
|
-
|
176
|
-
|
177
|
-
|
178
|
-
|
179
|
-
def training(loss_op):
|
180
|
-
|
181
|
-
with tf.name_scope("training") as scope:
|
182
|
-
|
183
|
-
training_op = optimizer.minimize(loss_op)
|
184
|
-
|
185
|
-
return training_op
|
186
|
-
|
187
|
-
|
188
|
-
|
189
|
-
|
190
|
-
|
191
|
-
def calc_accuracy(output_op, prints=False):
|
192
|
-
|
193
|
-
inputs, ts = make_prediction(num_of_prediction_epochs)
|
194
|
-
|
195
|
-
pred_dict = {
|
196
|
-
|
197
|
-
input_ph: inputs,
|
198
|
-
|
199
|
-
supervisor_ph: ts,
|
200
|
-
|
201
|
-
istate_ph: np.zeros((num_of_prediction_epochs, num_of_hidden_nodes * 2)),
|
202
|
-
|
203
|
-
}
|
204
|
-
|
205
|
-
output = sess.run([output_op], feed_dict=pred_dict)
|
206
|
-
|
207
|
-
|
208
|
-
|
209
|
-
def print_result(i, p, q):
|
210
|
-
|
211
|
-
[print(list(x)[0]) for x in i]
|
212
|
-
|
213
|
-
print("output: %f, correct: %d" % (p, q))
|
214
|
-
|
215
|
-
if prints:
|
216
|
-
|
217
|
-
[print_result(i, p, q) for i, p, q in zip(inputs, output[0], ts)]
|
218
|
-
|
219
|
-
|
220
|
-
|
221
|
-
opt = abs(output - ts)[0]
|
222
|
-
|
223
|
-
total = sum([1 if x[0] < 0.5 else 0 for x in opt]) ← 0.05から0.5に変更
|
224
|
-
|
225
|
-
print("accuracy %f" % (total / float(len(ts))))
|
226
|
-
|
227
|
-
return output
|
228
|
-
|
229
|
-
|
230
|
-
|
231
|
-
random.seed(0)
|
232
|
-
|
233
|
-
np.random.seed(0)
|
234
|
-
|
235
|
-
tf.set_random_seed(0)
|
236
|
-
|
237
|
-
|
238
|
-
|
239
|
-
optimizer = tf.train.GradientDescentOptimizer(learning_rate=learning_rate)
|
240
|
-
|
241
|
-
|
242
|
-
|
243
|
-
X, t = create_data(num_of_sample, length_of_sequences)
|
244
|
-
|
245
|
-
|
246
|
-
|
247
|
-
with tf.Graph().as_default():
|
248
|
-
|
249
|
-
input_ph = tf.placeholder(tf.float32, [None, length_of_sequences, num_of_input_nodes], name="input")
|
250
|
-
|
251
|
-
supervisor_ph = tf.placeholder(tf.float32, [None, num_of_output_nodes], name="supervisor")
|
252
|
-
|
253
|
-
istate_ph = tf.placeholder(tf.float32, [None, num_of_hidden_nodes * 2], name="istate")
|
254
|
-
|
255
|
-
|
256
|
-
|
257
|
-
output_op, states_op, datas_op = inference(input_ph, istate_ph)
|
258
|
-
|
259
|
-
loss_op = loss(output_op, supervisor_ph)
|
260
|
-
|
261
|
-
training_op = training(loss_op)
|
262
|
-
|
263
|
-
|
264
|
-
|
265
|
-
summary_op = tf.summary.merge_all()
|
266
|
-
|
267
|
-
init = tf.initialize_all_variables()
|
268
|
-
|
269
|
-
|
270
|
-
|
271
|
-
with tf.Session() as sess:
|
272
|
-
|
273
|
-
saver = tf.train.Saver()
|
274
|
-
|
275
|
-
summary_writer = tf.summary.FileWriter("/tmp/tensorflow_log", graph=sess.graph)
|
276
|
-
|
277
|
-
sess.run(init)
|
278
|
-
|
279
|
-
|
280
|
-
|
281
|
-
for epoch in range(num_of_training_epochs):
|
282
|
-
|
283
|
-
inputs, supervisors = get_batch(size_of_mini_batch, X, t)
|
284
|
-
|
285
|
-
train_dict = {
|
286
|
-
|
287
|
-
input_ph: inputs,
|
288
|
-
|
289
|
-
supervisor_ph: supervisors,
|
290
|
-
|
291
|
-
istate_ph: np.zeros((size_of_mini_batch, num_of_hidden_nodes * 2)),
|
292
|
-
|
293
|
-
}
|
294
|
-
|
295
|
-
sess.run(training_op, feed_dict=train_dict)
|
296
|
-
|
297
|
-
|
298
|
-
|
299
|
-
if (epoch) % 100 == 0:
|
300
|
-
|
301
|
-
summary_str, train_loss = sess.run([summary_op, loss_op], feed_dict=train_dict)
|
302
|
-
|
303
|
-
print("train#%d, train loss: %e" % (epoch, train_loss))
|
304
|
-
|
305
|
-
summary_writer.add_summary(summary_str, epoch)
|
306
|
-
|
307
|
-
if (epoch) % 500 == 0:
|
308
|
-
|
309
|
-
calc_accuracy(output_op)
|
310
|
-
|
311
|
-
|
312
|
-
|
313
|
-
calc_accuracy(output_op, prints=True)
|
314
|
-
|
315
|
-
datas = sess.run(datas_op)
|
316
|
-
|
317
|
-
saver.save(sess, "model.ckpt")
|
1
コードブロックの使用
test
CHANGED
File without changes
|
test
CHANGED
@@ -2,7 +2,11 @@
|
|
2
2
|
|
3
3
|
|
4
4
|
|
5
|
+
```
|
6
|
+
|
7
|
+
python
|
8
|
+
|
5
|
-
|
9
|
+
```で機械学習のプログラムを作るうえで
|
6
10
|
|
7
11
|
一変数ではなく、多変数のデータを読み込んで実行したい
|
8
12
|
|