Created
September 9, 2017 16:44
-
-
Save yampy/8bf645b893b64e91aeb87372a1d7a5f3 to your computer and use it in GitHub Desktop.
Keras/Tensorflowで始めるディープラーニングの基礎 ref: http://qiita.com/yampy/items/975f28594f295702ef18
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
pip install --upgrade tensorflow | |
pip install keras |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
from keras.models import Sequential | |
from keras.layers import Dense, Activation | |
# モデルの作成 | |
model = Sequential() | |
# モデルにレイヤーを積み上げていく | |
model.add(Dense(units=64, input_dim=100)) | |
model.add(Activation('relu')) | |
model.add(Dense(units=10)) | |
model.add(Activation('softmax')) | |
# 訓練プロセスの定義 | |
model.compile(loss='categorical_crossentropy', | |
optimizer='sgd', | |
metrics=['accuracy']) | |
# 訓練の実行 | |
# (x_train, y_trainはNumpy行列の学習データ) | |
model.fit(x_train, y_train, epochs=5, batch_size=32) | |
# 予測の実行 | |
classes = model.predict(x_test, batch_size=128) | |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Train on 60000 samples, validate on 10000 samples | |
Epoch 1/10 | |
59904/60000 [============================>.] - ETA: 0s - loss: 0.3267 - acc: 0.9009Epoch 00000: saving model to ./out/checkpoints/weights.00-0.08.hdf5 | |
60000/60000 [==============================] - 135s - loss: 0.3263 - acc: 0.9011 - val_loss: 0.0785 - val_acc: 0.9749 | |
Epoch 2/10 | |
59904/60000 [============================>.] - ETA: 0s - loss: 0.1132 - acc: 0.9671Epoch 00001: saving model to ./out/checkpoints/weights.01-0.06.hdf5 | |
60000/60000 [==============================] - 131s - loss: 0.1131 - acc: 0.9671 - val_loss: 0.0580 - val_acc: 0.9811 | |
Epoch 3/10 | |
59904/60000 [============================>.] - ETA: 0s - loss: 0.0871 - acc: 0.9735Epoch 00002: saving model to ./out/checkpoints/weights.02-0.05.hdf5 | |
60000/60000 [==============================] - 135s - loss: 0.0870 - acc: 0.9735 - val_loss: 0.0469 - val_acc: 0.9850 | |
(中略) | |
Epoch 10/10 | |
59904/60000 [============================>.] - ETA: 0s - loss: 0.0410 - acc: 0.9876Epoch 00009: saving model to ./out/checkpoints/weights.09-0.03.hdf5 | |
60000/60000 [==============================] - 140s - loss: 0.0410 - acc: 0.9876 - val_loss: 0.0295 - val_acc: 0.9902 |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
tensorboard --logdir=./out/ |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
from keras.layers import Input, Dense | |
from keras.models import Model | |
# インプットの定義 | |
inputs = Input(shape=(784,)) | |
# レイヤーの定義 | |
nw = Dense(64, activation='relu')(inputs) | |
nw = Dense(64, activation='relu')(x) | |
predictions = Dense(10, activation='softmax')(x) | |
# モデルの定義(インプットとレイヤーを指定) | |
model = Model(inputs=inputs, outputs=predictions) | |
# 訓練プロセスの定義 | |
model.compile(optimizer='rmsprop', | |
loss='categorical_crossentropy', | |
metrics=['accuracy']) | |
# 訓練の実行 | |
model.fit(data, labels) | |
# 予測の実行 | |
classes = model.predict(x_test, batch_size=128) | |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Train on 60000 samples, validate on 10000 samples | |
Epoch 1/20 | |
60000/60000 [==============================] - 7s - loss: 0.3319 - acc: 0.8974 - val_loss: 0.1267 - val_acc: 0.9607 | |
Epoch 2/20 | |
60000/60000 [==============================] - 6s - loss: 0.1621 - acc: 0.9524 - val_loss: 0.0921 - val_acc: 0.9733 | |
Epoch 3/20 | |
60000/60000 [==============================] - 6s - loss: 0.1318 - acc: 0.9610 - val_loss: 0.0874 - val_acc: 0.9750 | |
(中略) | |
Epoch 20/20 | |
60000/60000 [==============================] - 6s - loss: 0.0722 - acc: 0.9823 - val_loss: 0.1083 - val_acc: 0.9817 |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment