Question

我想在keras中编写一个自定义损失函数，该函数还使用输出与输入的梯度

import numpy as np
from keras.models import Sequential
from keras.layers import Dense, Dropout

model = Sequential()
model.add(Dense(200, input_dim=20, activation='relu'))
model.add(Dense(64, input_dim=20, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(64, activation='relu'))
model.add(Dropout(0.5))
model.add(Dense(1, activation='sigmoid'))
model.compile(loss='binary_crossentropy',
          optimizer='rmsprop',
          metrics=['accuracy'])

现在，我要替换损失函数。我只知道如何使用tensorflow编写如下：

x = tf.placeholder(tf.float32, shape=[None, 1], name="x") 
ext_f0 = tf.placeholder(tf.float32, shape=[None, 1], name="f") 
xx0 = tf.concat([[[R_variable['x_start'] * 1.0],[R_variable['x_end'] * 1.0]], x], 0)
yy0 = univAprox(xx0)
Boundary_y = yy0[0:2]
y = yy0[2:]
GradU = tf.gradients(y,x)  #***KEY***
GradSum = tf.reduce_sum(tf.square(GradU)) / 2 * dx_train
FSum = tf.reduce_sum(tf.multiply(ext_f0,y)) * dx_train
loss = GradSum + FSum + Beta * tf.reduce_sum(tf.square(Boundary_y))

关键是我需要获得GradU。我该怎么做在喀拉拉邦？

Answer 1

您可以编写自定义损失函数。您应该知道的是它的形式：

def custom_loss(y_true, y_pred):
  # do something to compute loss
  ...
  return loss

就应该这样。

自定义keras损失函数

1 个答案: