Question

我有以下代码，可以并行训练具有不同数据集的相同模型。我想知道在此代码中得到的结果与如果使用相应的数据集对模型进行一次训练之间有什么区别。非常感谢您的帮助。

from keras.layers import *
from keras.models import Model, Sequential
import numpy as np
import random
import test
index=[]
for i in range(10):
    index.append(random.sample(range(0, 201), 10))
x_tr=[]
y_tr=[]
x_te=[]
y_te=[]

aa = test.train_data(index[0])
X_train0, Y_train0, X_Test0, Y_Test0 = test.ddata(aa)
aa = test.train_data(index[1])
X_train1, Y_train1, X_Test1, Y_Test1 = test.ddata(aa)
aa = test.train_data(index[2])
X_train2, Y_train2, X_Test2, Y_Test2 = test.ddata(aa)
aa = test.train_data(index[3])
X_train3, Y_train3, X_Test3, Y_Test3 = test.ddata(aa)
aa = test.train_data(index[4])
X_train4, Y_train4, X_Test4, Y_Test4 = test.ddata(aa)
aa = test.train_data(index[5])
X_train5, Y_train5, X_Test5, Y_Test5 = test.ddata(aa)
aa = test.train_data(index[6])
X_train6, Y_train6, X_Test6,Y_Test6 = test.ddata(aa)
aa = test.train_data(index[7])
X_train7, Y_train7, X_Test7, Y_Test7 = test.ddata(aa)
aa = test.train_data(index[8])
X_train8, Y_train8, X_Test8, Y_Test8 = test.ddata(aa)
aa = test.train_data(index[9])
X_train9, Y_train9, X_Test9, Y_Test9 = test.ddata(aa)


m=test.get_model()
inp0=Input((5,10,10,1))
inp1=Input((5,10,10,1))
inp2=Input((5,10,10,1))
inp3=Input((5,10,10,1))
inp4=Input((5,10,10,1))
inp5=Input((5,10,10,1))
inp6=Input((5,10,10,1))
inp7=Input((5,10,10,1))
inp8=Input((5,10,10,1))
inp9=Input((5,10,10,1))
out0=m(inp0)
out1=m(inp1)
out2=m(inp2)
out3=m(inp3)
out4=m(inp4)
out5=m(inp5)
out6=m(inp6)
out7=m(inp7)
out8=m(inp8)
out9=m(inp9)





model = Model([inp0,inp1,inp2,inp3,inp4,inp5,inp6,inp7,inp8,inp9],[out0,out1,out2,out3,out4,out5,out6,out7,out8,out9])
model.compile(optimizer='adam', loss='mse')
model.fit([X_train0,X_train1,X_train2,X_train3,X_train4,X_train5,X_train6,X_train7,X_train8,X_train9],[Y_train0,Y_train1,Y_train2,Y_train3,Y_train4,Y_train5,Y_train6,Y_train7,Y_train8,Y_train9], epochs = 50)
ypred0,ypred1,ypred2,ypred3,ypred4,ypred5,ypred6,ypred7,ypred8,ypred9 = model.predict([X_Test0,X_Test1,X_Test2,X_Test3,X_Test4,X_Test5,X_Test6,X_Test7,X_Test8,X_Test9])
print(ypred0.shape)

FYI测试是我从中获取数据的另一个地方。

Answer 1

最好是并行训练或一次训练而不是顺序训练。

如果按顺序训练它，最终可能会过度拟合训练模型的最后一个数据集。

在keras与顺序中训练并行模型

1 个答案: