<code>from keras.utils import plot_model

import pandas as pd

import numpy as np

import tensorflow as tf

import matplotlib.pyplot as plt

def get_activations(model, inputs, print_shape_only=False, layer_name=None):

# From: https://github.com/philipperemy/keras-visualize-activations

activations = []

inp = model.input

if layer_name is None:

outputs = [layer.output for layer in model.layers]

else:

outputs = [layer.output for layer in model.layers if layer.name == layer_name] # all layer outputs

funcs = [K.function([inp] + [K.learning_phase()], [out]) for out in outputs] # evaluation functions

layer_outputs = [func([inputs, 1.])[0] for func in funcs]

for layer_activations in layer_outputs:

activations.append(layer_activations)

if print_shape_only:

print(layer_activations.shape)

else:

print(layer_activations.shape)

print(layer_activations)

return activations

def get_data(n, input_dim, attention_column=1):

"""

:param n: the number of samples to retrieve.

:param input_dim: the number of dimensions of each element in the series.

:param attention_column: the column linked to the target. Everything else is purely random.

:return: x: model inputs, y: model targets

"""

x = np.random.standard_normal(size=(n, input_dim))

y = np.random.randint(low=0, high=2, size=(n, 1))

x[:, attention_column] = y[:, 0]

return x, y

def get_data_recurrent(n, time_steps, input_dim, attention_column=10):

"""

:param n: the number of samples to retrieve.

:param time_steps: the number of time steps of your series.

:param input_dim: the number of dimensions of each element in the series.

:param attention_column: the column linked to the target. Everything else is purely random.

:return: x: model inputs, y: model targets

"""

x = np.random.standard_normal(size=(n, time_steps, input_dim))

y = np.random.randint(low=0, high=2, size=(n, 1))

x[:, attention_column, :] = np.tile(y[:], (1, input_dim))

return x, y

def get_data_recurrent2(n, time_steps, input_dim, attention_dim=5):

x = np.random.standard_normal(size=(n, time_steps, input_dim))

y = np.random.randint(low=0, high=2, size=(n, 1))

x[:,:,attention_dim] = np.tile(y[:], (1, time_steps))

return x,y

from keras.layers import Input, Dense, LSTM ,Conv1D,Dropout,Bidirectional,Multiply

from keras.models import Model

#from attention_utils import get_activations

from keras.layers import Multiply

from keras.layers.core import *

from keras.layers import LSTM

from keras.models import *

SINGLE_ATTENTION_VECTOR = False

def attention_3d_block(inputs):

# inputs.shape = (batch_size, time_steps, input_dim)

input_dim = int(inputs.shape[2])

a = inputs

a = Dense(input_dim, activation='softmax')(a)code>

if SINGLE_ATTENTION_VECTOR:

a = Lambda(lambda x: K.mean(x, axis=1), name='dim_reduction')(a)code>

a = RepeatVector(input_dim)(a)

a_probs = Permute((1, 2), name='attention_vec')(a)code>

#output_attention_mul = merge([inputs, a_probs], name='attention_mul', mode='mul')code>

output_attention_mul = Multiply()([inputs, a_probs])

return output_attention_mul

# Another way of writing the attention mechanism is suitable for the use of the above error source:https://blog.csdn.net/uhauha2929/article/details/80733255

def attention_3d_block2(inputs, single_attention_vector=False):

# If the upper layer is LSTM, you need return_sequences=True

# inputs.shape = (batch_size, time_steps, input_dim)

time_steps = K.int_shape(inputs)[1]

input_dim = K.int_shape(inputs)[2]

a = Permute((2, 1))(inputs)

a = Dense(time_steps, activation='softmax')(a)code>

if single_attention_vector:

a = Lambda(lambda x: K.mean(x, axis=1))(a)

a = RepeatVector(input_dim)(a)

a_probs = Permute((2, 1))(a)

# Multiplied by the attention weight, but there is no summation, it seems to have little effect

# If you classify tasks, you can do Flatten expansion

# element-wise

output_attention_mul = Multiply()([inputs, a_probs])

return output_attention_mul

如需完整代码，三连私信