Question

我在创建具有3个层（激活函数为cRelu）的Dense模型时遇到问题。 cRelu连接两个relu（一个负数和一个正数）并在其输出中创建两倍大小的张量。尝试在其后添加另一层时，总是出现尺寸不匹配错误

model  = Sequential()
model.add(Dense(N, input_dim=K, activation=crelu))
model.add(Dense(N//2, activation=crelu))

如何告诉下一层期望输入2N并达到N？

Answer 1

Keras不希望激活功能会更改输出形状。如果要更改它，则应将crelu功能包装在一层中并指定相应的输出形状：

import ImagePicker from 'react-native-image-picker';

// More info on all the options is below in the API Reference... just some common use cases shown here
const options = {
  title: 'Select Avatar',
  customButtons: [{ name: 'fb', title: 'Choose Photo from Facebook' }],
  storageOptions: {
    skipBackup: true,
    path: 'images',
  },
};

/**
 * The first arg is the options object for customization (it can also be null or omitted for default options),
 * The second arg is the callback which sends object: response (more info in the API Reference)
 */
ImagePicker.showImagePicker(options, (response) => {
  console.log('Response = ', response);

  if (response.didCancel) {
    console.log('User cancelled image picker');
  } else if (response.error) {
    console.log('ImagePicker Error: ', response.error);
  } else if (response.customButton) {
    console.log('User tapped custom button: ', response.customButton);
  } else {
    const source = { uri: response.uri };

    // You can also display the image using data:
    // const source = { uri: 'data:image/jpeg;base64,' + response.data };

    this.setState({
      avatarSource: source,
    });
  }
});

然后您可以按以下方式使用它

import tensorflow as tf
from keras.layers import Layer

class cRelu(Layer):

    def __init__(self, **kwargs):
        super(cRelu, self).__init__(**kwargs)

    def build(self, input_shape):
        super(cRelu, self).build(input_shape)

    def call(self, x):
        return tf.nn.crelu(x)

    def compute_output_shape(self, input_shape):
        """
        All axis of output_shape, except the last one,
        coincide with the input shape.
        The last one is twice the size of the corresponding input 
        as it's the axis along which the two relu get concatenated.
        """
        return (*input_shape[:-1], input_shape[-1]*2)

具有cRelu激活的Keras序列模型

1 个答案: