怠け者のためのディープラーニング入門 - Scikit-learn・Keras ref: http://qiita.com/akimach/items/e9088ab1b98a06ae9454

file0.txt

pip3 install scikit-learn

file1.py

from sklearn import svm
X_train = [[0, 0], [1, 1]]
y_train = [0, 1]
X_test = [[1, 1], [0, 1]]
y_test = [1, 0]
clf = svm.SVC(C=1.0)
clf.fit(X_train, y_train)
acc = clf.score(X_test, y_test)
print("Accuracy:", acc)
X_new = [[0, 0], [0, 1]]
pred = clf.predict(X_new)
print(pred)

file2.txt

Accuracy: 0.5
Prediction result: [0 1]

file3.py

import numpy as np
from sklearn.utils import shuffle
from sklearn.model_selection import train_test_split
np.random.seed(1234)
X = np.random.randint(10, size=(10, 4))
y = np.random.randint(4, size=(10))
# Randomly permute
X, y = shuffle(X, y)
# Split dataset
# Train : Test = 8 : 2
X_train, X_test, y_train, y_test =\
    train_test_split(X, y, test_size=0.2)
print("Train:", X_train.shape, y_train.shape)
print("Test:", X_test.shape, y_test.shape)

file4.py

Train: (8, 4) (8,)
Test: (2, 4) (2,)

file5.txt

pip3 install keras

file7.txt

Using TensorFlow backend.
Epoch 1/10
160/160 [==============================] - 0s - loss: 0.6587 - acc: 0.7938
Epoch 2/10
160/160 [==============================] - 0s - loss: 0.6303 - acc: 0.8625
Epoch 3/10
160/160 [==============================] - 0s - loss: 0.6040 - acc: 0.8625
Epoch 4/10
160/160 [==============================] - 0s - loss: 0.5795 - acc: 0.8812
Epoch 5/10
160/160 [==============================] - 0s - loss: 0.5562 - acc: 0.8875
Epoch 6/10
160/160 [==============================] - 0s - loss: 0.5345 - acc: 0.8875
Epoch 7/10
160/160 [==============================] - 0s - loss: 0.5143 - acc: 0.8937
Epoch 8/10
160/160 [==============================] - 0s - loss: 0.4950 - acc: 0.9000
Epoch 9/10
160/160 [==============================] - 0s - loss: 0.4770 - acc: 0.9125
Epoch 10/10
160/160 [==============================] - 0s - loss: 0.4599 - acc: 0.9125
32/40 [=======================>......] - ETA: 0s
Loss: 0.464164888859 Accuracy: 0.9
[[ 0.57916182]
 [ 0.50237983]
 [ 0.70500255]
 [ 0.63764608]
 [ 0.33048654]]
[[ 0.42043626]
 [ 0.32738858]
 [ 0.27511841]
 [ 0.32091939]
 [ 0.32354838]]

mlp.py

import numpy as np
from keras.models import Sequential
from keras.layers import Dense, Activation
from keras import optimizers
from sklearn.utils import shuffle
from sklearn.model_selection import train_test_split

# Setup dataset
np.random.seed(1234)
n_data = 200
n_class = n_data//2
X_positive = np.random.normal(loc=1.0, scale=1.0, size=(n_class, 2))
X_negative = np.random.normal(loc=-1.0, scale=1.0, size=(n_class, 2))
X = np.r_[X_positive, X_negative]
y = np.r_[[1]*n_class, [0]*n_class]
# Split dataset
X, y = shuffle(X, y)
X_train, X_test, y_train, y_test =\
        train_test_split(X, y, test_size=0.2)
# Build a model
model = Sequential()
model.add(Dense(units=128, input_dim=2))
model.add(Activation('relu'))
model.add(Dense(units=128))
model.add(Activation('relu'))
model.add(Dense(units=1))
model.add(Activation('sigmoid'))
model.compile(loss='binary_crossentropy',
              optimizer=optimizers.SGD(lr=0.01),
              metrics=['accuracy',])
# Training
model.fit(X_train, y_train, batch_size=20, epochs=10)
# Evaluation
score = model.evaluate(X_test, y_test)
print("\nLoss:", score[0], "Accuracy:", score[1])
# Prediction
X_positive_new = np.random.normal(loc=1.0, scale=1.0, size=(5, 2))
X_negative_new = np.random.normal(loc=-1.0, scale=1.0, size=(5, 2))
print(model.predict(X_positive_new))
print(model.predict(X_negative_new))

plot_mlp.py

# mlp.pyに追加してください
# ... 略 ...
from keras.utils import plot_model
# ... 略 ...
from keras.utils import plot_model
plot_model(model, to_file='model.png')
"""
Jupyter Notebook上で可視化する場合は以下のようにする : 
from IPython.display import SVG
g = model_to_dot(model=m)
SVG(g.create(prog='dot', format='svg'))
"""