keras 自定义loss层+接受输入实例

时间:2022-09-04 17:36:20

loss函数如何接受输入

keras封装的比较厉害,官网给的例子写的云里雾里,

在*找到了答案

You can wrap the loss function as a inner function and pass your input tensor to it (as commonly done when passing additional arguments to the loss function).

?
1
2
3
4
def custom_loss_wrapper(input_tensor):
 def custom_loss(y_true, y_pred):
  return K.binary_crossentropy(y_true, y_pred) + K.mean(input_tensor)
 return custom_loss
?
1
2
3
4
5
input_tensor = Input(shape=(10,))
hidden = Dense(100, activation='relu')(input_tensor)
out = Dense(1, activation='sigmoid')(hidden)
model = Model(input_tensor, out)
model.compile(loss=custom_loss_wrapper(input_tensor), optimizer='adam')

You can verify that input_tensor and the loss value will change as different X is passed to the model.

?
1
2
3
4
5
X = np.random.rand(1000, 10)
y = np.random.randint(2, size=1000)
model.test_on_batch(X, y) # => 1.1974642
X *= 1000
model.test_on_batch(X, y) # => 511.15466

fit_generator

fit_generator ultimately calls train_on_batch which allows for x to be a dictionary.

Also, it could be a list, in which casex is expected to map 1:1 to the inputs defined in Model(input=[in1, …], …)

?
1
2
3
4
### generator
yield [inputX_1,inputX_2],y
### model
model = Model(inputs=[inputX_1,inputX_2],outputs=...)

补充知识:keras中自定义 loss损失函数和修改不同样本的loss权重(样本权重、类别权重)

首先辨析一下概念:

1. loss是整体网络进行优化的目标, 是需要参与到优化运算,更新权值W的过程的

2. metric只是作为评价网络表现的一种“指标”, 比如accuracy,是为了直观地了解算法的效果,充当view的作用,并不参与到优化过程

一、keras自定义损失函数

在keras中实现自定义loss, 可以有两种方式,一种自定义 loss function, 例如:

?
1
2
3
4
5
6
7
# 方式一
def vae_loss(x, x_decoded_mean):
 xent_loss = objectives.binary_crossentropy(x, x_decoded_mean)
 kl_loss = - 0.5 * K.mean(1 + z_log_sigma - K.square(z_mean) - K.exp(z_log_sigma), axis=-1)
 return xent_loss + kl_loss
 
vae.compile(optimizer='rmsprop', loss=vae_loss)

或者通过自定义一个keras的层(layer)来达到目的, 作为model的最后一层,最后令model.compile中的loss=None:

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
# 方式二
# Custom loss layer
class CustomVariationalLayer(Layer):
 
 def __init__(self, **kwargs):
  self.is_placeholder = True
  super(CustomVariationalLayer, self).__init__(**kwargs)
 def vae_loss(self, x, x_decoded_mean_squash):
 
  x = K.flatten(x)
  x_decoded_mean_squash = K.flatten(x_decoded_mean_squash)
  xent_loss = img_rows * img_cols * metrics.binary_crossentropy(x, x_decoded_mean_squash)
  kl_loss = - 0.5 * K.mean(1 + z_log_var - K.square(z_mean) - K.exp(z_log_var), axis=-1)
  return K.mean(xent_loss + kl_loss)
 
 def call(self, inputs):
 
  x = inputs[0]
  x_decoded_mean_squash = inputs[1]
  loss = self.vae_loss(x, x_decoded_mean_squash)
  self.add_loss(loss, inputs=inputs)
  # We don't use this output.
  return x
 
y = CustomVariationalLayer()([x, x_decoded_mean_squash])
vae = Model(x, y)
vae.compile(optimizer='rmsprop', loss=None)

在keras中自定义metric非常简单,需要用y_pred和y_true作为自定义metric函数的输入参数 点击查看metric的设置

注意事项:

1. keras中定义loss,返回的是batch_size长度的tensor, 而不是像tensorflow中那样是一个scalar

2. 为了能够将自定义的loss保存到model, 以及可以之后能够顺利load model, 需要把自定义的loss拷贝到keras.losses.py 源代码文件下,否则运行时找不到相关信息,keras会报错

有时需要不同的sample的loss施加不同的权重,这时需要用到sample_weight,例如

discriminator.train_on_batch(imgs, [valid, labels], class_weight=class_weights)

二、keras中的样本权重

?
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
# Import
import numpy as np
from sklearn.utils import class_weight
 
# Example model
model = Sequential()
model.add(Dense(32, activation='relu', input_dim=100))
model.add(Dense(1, activation='sigmoid'))
 
# Use binary crossentropy loss
model.compile(optimizer='rmsprop',
    loss='binary_crossentropy',
    metrics=['accuracy'])
 
# Calculate the weights for each class so that we can balance the data
weights = class_weight.compute_class_weight('balanced',
           np.unique(y_train),
           y_train)
 
# Add the class weights to the training          
model.fit(x_train, y_train, epochs=10, batch_size=32, class_weight=weights)

Note that the output of the class_weight.compute_class_weight() is an numpy array like this: [2.57569845 0.68250928].

以上这篇keras 自定义loss层+接受输入实例就是小编分享给大家的全部内容了,希望能给大家一个参考,也希望大家多多支持服务器之家。

原文链接:https://blog.csdn.net/u013608336/article/details/82559469