静态与动态RNN比较
Tensorflow中提供了两种创建RNN结构的函数:
- tf.nn.static_rnn
- tf.nn.dynamic_rnn
两个函数的区别如下[1]:
tf.nn.static_rnn creates an unrolled graph for a fixed RNN length. That means, if you call tf.nn.static_rnn with inputs having 200 time steps you are creating a static graph with 200 RNN steps. First, graph creation is slow. Second, you’re unable to pass in longer sequences (> 200) than you’ve originally specified.
tf.nn.dynamic_rnn solves this. It uses a tf.While loop to dynamically construct the graph when it is executed. That means graph creation is faster and you can feed batches of variable size.
import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data
file_dir = '/Users/wangruidong/Documents/MachineLearning/Dataset/MNIST/'
mnist = input_data.read_data_sets(file_dir, one_hot=True)
train_img = mnist.train.images
train_labels = mnist.train.labels
print(train_img.shape)
print(train_labels.shape)
print('Load data finish.')
Extracting /Users/wangruidong/Documents/MachineLearning/Dataset/MNIST/train-images-idx3-ubyte.gz
Extracting /Users/wangruidong/Documents/MachineLearning/Dataset/MNIST/train-labels-idx1-ubyte.gz
Extracting /Users/wangruidong/Documents/MachineLearning/Dataset/MNIST/t10k-images-idx3-ubyte.gz
Extracting /Users/wangruidong/Documents/MachineLearning/Dataset/MNIST/t10k-labels-idx1-ubyte.gz
(55000, 784)
(55000, 10)
Load data finish.
input_dim = 28
time_step = 28