在keras下實現多個模型的融合
小風風12580 2019-09-30 10:42:00 1105 收藏 7
展開
在網上搜過發現關於keras下的模型融合框架其實很簡單，奈何網上說了一大堆，這個東西官方文檔上就有，自己寫了個demo：
# Function：基於keras框架下實現，多個獨立任務分類
# Writer: PQF
# Time: 2019/9/29
import numpy as np
from keras.layers import Input, Dense
from keras.models import Model
import tensorflow as tf
# 生成訓練集
dataset_size = 128*3
rdm = np.random.RandomState(1)
X = rdm.rand(dataset_size,2)
Y1 = [[int(x1+x2<1)] for (x1,x2) in X]
Y2 = [[int(x1+x2*x2<0.5)] for (x1,x2) in X]
X_train = X[:-2]
Y_train1 = Y1[:-2]
Y_train2 = Y2[:-2]
X_test = X[-2:dataset_size]
Y_test1 = Y1[-2:dataset_size]
Y_test2 = Y2[-2:dataset_size]
#網絡一
input = Input(shape=(2,))
x = Dense(units=16,activation='relu')(input)
output = Dense(units=1,activation='sigmoid',name='output1')(x)
#網絡二
input2 = Input(shape=(2,))
x2 = Dense(units=16,activation='relu')(input2)
output2 = Dense(units=1,activation='sigmoid',name='output2')(x2)
#模型合並
model = Model(inputs=[input,input2],outputs=[output,output2])
model.summary()
model.compile(optimizer='rmsprop',loss='binary_crossentropy',loss_weights=[1.0,1.0])
model.fit([X_train,X_train],[Y_train1,Y_train2],batch_size=48,epochs=200)
print('x_test is :\n')
print(X_test)
print('y_test1 is :\n')
print(Y_test1)
print('y_test2 is :\n')
print(Y_test2)
predict = model.predict([X_test,X_test])
print('prediction is : \n')
print(predict[0])
print(predict[1])
————————————————
版權聲明：本文為CSDN博主「小風風12580」的原創文章，遵循CC 4.0 BY-SA版權協議，轉載請附上原文出處鏈接及本聲明。
原文鏈接：https://blog.csdn.net/weixin_43392276/java/article/details/101757173