如何在Keras中使用精度(而不是精度)优化CNN

Question

我的问题很简单，我有一个具有True和False值的目标列。基本上，这是一个二元分类问题。我想知道如何使用Precision作为度量而不是Accuracy来优化我的CNN ?

Btw，这是不工作:

model.compile(loss='binary_crossentropy',  optimizer=optm, metrics=['precision'])

这是我的代码:

model = Sequential()
model.add(Dense(64,name = 'Primera', input_dim=8, activation='relu'))
model.add(Dense(32 ,name = 'Segunda'))
model.add(Dense(1,name = 'Tercera', activation='sigmoid'))
from tensorflow.keras import optimizers
optm = optimizers.Adam(learning_rate=0.001, beta_1=0.9, beta_2=0.999, amsgrad=False)
model.compile(loss='binary_crossentropy',  optimizer=optm, metrics=['accuracy'])
model.summary()
history = model.fit(trainX, trainY, 
epochs=1000, 
batch_size=16, 
validation_split=0.1, 
verbose=1)

Answer 1

您可以使用tf.keras.metrics.Precision()，参见下面的示例代码。

from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense
from tensorflow.keras.optimizers import Adam
from tensorflow.keras.metrics import Precision
from sklearn.datasets import make_classification
X, y = make_classification(n_classes=2, n_features=8, n_informative=8, n_redundant=0, random_state=42)
model = Sequential()
model.add(Dense(64, input_dim=8, activation='relu'))
model.add(Dense(32))
model.add(Dense(1, activation='sigmoid'))
model.compile(
loss='binary_crossentropy',
optimizer=Adam(learning_rate=0.001, beta_1=0.9, beta_2=0.999, amsgrad=False),
metrics=[Precision()]
)
model.fit(X, y, epochs=5, batch_size=32, validation_split=0.1, verbose=1)
# Epoch 1/5
# 3/3 [==============================] - 1s 83ms/step - loss: 0.8535 - precision: 0.5116 - val_loss: 0.6936 - val_precision: 0.5714
# Epoch 2/5
# 3/3 [==============================] - 0s 7ms/step - loss: 0.6851 - precision: 0.5200 - val_loss: 0.5975 - val_precision: 0.6667
# Epoch 3/5
# 3/3 [==============================] - 0s 7ms/step - loss: 0.6004 - precision: 0.6545 - val_loss: 0.5370 - val_precision: 0.8000
# Epoch 4/5
# 3/3 [==============================] - 0s 7ms/step - loss: 0.5412 - precision: 0.8250 - val_loss: 0.4878 - val_precision: 0.8000
# Epoch 5/5
# 3/3 [==============================] - 0s 8ms/step - loss: 0.5145 - precision: 0.9394 - val_loss: 0.4462 - val_precision: 0.8000