DecafTea的博客

理解encoder_outputs, state_h, state_c = encoder(encoder_inputs)的三个输出：https://huhuhang.com/post/machine-learning/lstm-return-sequences-stateinputs = tf.keras.layers.Input(shape=(3, 1))lstm = tf.keras.layers.LSTM(1, return_state=True)(inputs)model = tf.ke

1. 定义模型架构序贯式和函数式都可以定义模型架构，序贯式是层层叠加定义模型，函数式是通过input-middle_output1-middle_output2-…-output定义模型。一般两种方式可以互换。序贯式：函数式：但是，对于inception network这种input-多个middle_output-output这种需要“分流”的情况，只能用函数式定义模型。Note：Input(shape=())定义输入的dimkeras.layers.concatenate(

input, noise = {ndarry: }input.size = {tuple: }希望预测出的非线性函数：x^2generate 200 training examples：（x，x^2+noise), x between -0.5 and 0.5import numpy as npimport kerasimport matplotlib.pyplot as pltfrom keras.models import Sequentialfrom keras.layers imp