如何为神经网络模型创建管道？

数据挖掘 神经网络 深度学习
2022-02-11 02:43:18

我尝试过创建这样的管道，但出现错误

- AttributeError：“管道”对象没有“编译”属性

scaler = StandardScaler()
model = Sequential()

model.add(Dense(120, input_dim=46,activation='relu'))
model.add(Dropout(0.1, noise_shape=None, seed=None))

model.add(Dense(80, activation='relu'))
model.add(Dropout(0.1, noise_shape=None, seed=None))

model.add(Dense(40, activation='relu'))
model.add(Dense(1, activation='sigmoid'))

pipeline = make_pipeline(scaler,model)
pipeline.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])


pipeline.fit(X_train,y_train, epochs=50, batch_size=20, validation_data = (X_test,y_test))

# evaluate the model
scores = pipeline.evaluate(X_test,y_test)
print("\n%s: %.2f%%" % (model.metrics_names[1], scores[1]*100))
2个回答

我可以用下面的代码来实现这一点

谢谢@JahKnows

scaler = StandardScaler()
model = Sequential()

model.add(Dense(120, input_dim=46,activation='relu'))
model.add(Dropout(0.1, noise_shape=None, seed=None))

model.add(Dense(80, activation='relu'))
model.add(Dropout(0.1, noise_shape=None, seed=None))

model.add(Dense(40, activation='relu'))
model.add(Dense(1, activation='sigmoid'))
model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])

pipeline = Pipeline([("scaler",scaler),("model",model)])
pipeline.fit(X_train,y_train, model__epochs=50, model__batch_size=20, model__validation_data = (X_test,y_test))

# evaluate the model
nn_pred = grid_search.predict(X_test)

print('Accuracy of NN on test is:',accuracy_score(y_test,nn_pred))
print("\n%s: %.2f%%" % (model.metrics_names[1], scores[1]*100))

您应该在将模型放入管道之前对其进行编译。

scaler = StandardScaler()
model = Sequential()

model.add(Dense(120, input_dim=46,activation='relu'))
model.add(Dropout(0.1, noise_shape=None, seed=None))

model.add(Dense(80, activation='relu'))
model.add(Dropout(0.1, noise_shape=None, seed=None))

model.add(Dense(40, activation='relu'))
model.add(Dense(1, activation='sigmoid'))
model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])

pipeline = make_pipeline(scaler,model)
pipeline.fit(X_train,y_train, epochs=50, batch_size=20, validation_data = (X_test,y_test))

# evaluate the model
scores = pipeline.evaluate(X_test,y_test)
print("\n%s: %.2f%%" % (model.metrics_names[1], scores[1]*100))
其它你可能感兴趣的问题
上一篇知道垃圾邮件发送者的电子邮件地址将如何改进垃圾邮件检测算法？ 下一篇2018 年的强化学习，最佳技巧和窍门？