&引用；扫描"；theano和tensorflow中的函数

我有一个用theano编写的函数：

import numpy as np
import theano as theano
import theano.tensor as T

def forward_prop_step(x_t, s_t1_prev, s_t2_prev):
            # This is how we calculated the hidden state in a simple RNN. No longer!
            # s_t = T.tanh(U[:,x_t] + W.dot(s_t1_prev))

            # Word embedding layer
            x_e = E[:,x_t]

            # GRU Layer 1
            z_t1 = T.nnet.hard_sigmoid(U[0].dot(x_e) + W[0].dot(s_t1_prev) + b[0])
            r_t1 = T.nnet.hard_sigmoid(U[1].dot(x_e) + W[1].dot(s_t1_prev) + b[1])
            c_t1 = T.tanh(U[2].dot(x_e) + W[2].dot(s_t1_prev * r_t1) + b[2])
            s_t1 = (T.ones_like(z_t1) - z_t1) * c_t1 + z_t1 * s_t1_prev

            # GRU Layer 2
            z_t2 = T.nnet.hard_sigmoid(U[3].dot(s_t1) + W[3].dot(s_t2_prev) + b[3])
            r_t2 = T.nnet.hard_sigmoid(U[4].dot(s_t1) + W[4].dot(s_t2_prev) + b[4])
            c_t2 = T.tanh(U[5].dot(s_t1) + W[5].dot(s_t2_prev * r_t2) + b[5])
            s_t2 = (T.ones_like(z_t2) - z_t2) * c_t2 + z_t2 * s_t2_prev

            # Final output calculation
            # Theano's softmax returns a matrix with one row, we only need the row
            o_t = T.nnet.softmax(V.dot(s_t2) + c)[0]

return [o_t, s_t1, s_t2]

我用tensorflow重写了它：

import numpy as np, tensorflow as tf, operator

def forward_prop_step(x_t, s_t1_prev, s_t2_prev):
                        # This is how we calculated the hidden state in a simple RNN. No longer!
                        # s_t = T.tanh(U[:,x_t] + W.dot(s_t1_prev))

                        # Word embedding layer
                        x_e = E[:,x_t]

                        # GRU Layer 1
                        z_t1 = tf.nn.sigmoid(tf.reduce_sum(tf.multiply(U[0], x_e)) + tf.reduce_sum(tf.multiply(W[0], s_t1_prev)) + b[0])
                        r_t1 = tf.nn.sigmoid(tf.reduce_sum(tf.multiply(U[1], x_e)) + tf.reduce_sum(tf.multiply(W[1], s_t1_prev)) + b[1])
                        c_t1 = tf.nn.tanh(tf.reduce_sum(tf.multiply(U[2], x_e)) + tf.reduce_sum(tf.multiply(W[2], (s_t1_prev * r_t1))) + b[2])
                        s_t1 = (tf.ones_like(z_t1) - z_t1) * c_t1 + z_t1 * s_t1_prev

                        # GRU Layer 2
                        z_t2 = tf.nn.sigmoid(tf.reduce_sum(tf.multiply(U[3], s_t1)) + tf.reduce_sum(tf.multiply(W[3], s_t2_prev)) + b[3])
                        r_t2 = tf.nn.sigmoid(tf.reduce_sum(tf.multiply(U[4], s_t1)) + tf.reduce_sum(tf.multiply(W[4], s_t2_prev)) + b[4])
                        c_t2 = tf.nn.tanh(tf.reduce_sum(tf.multiply(U[5], s_t1)) + tf.reduce_sum(tf.multiply(W[5], (s_t2_prev * r_t2))) + b[5])
                        s_t2 = (tf.ones_like(z_t2) - z_t2) * c_t2 + z_t2 * s_t2_prev

                        # Final output calculation
                        # Tensorflow's softmax returns a matrix with one row, we only need the row
                        o_t = tf.nn.softmax(tf.reduce_sum(tf.multiply(V, s_t2)) + c)[0]

                        return [o_t, s_t1, s_t2]

在theano中，调用scan函数在循环中执行“forward\u prep\u step”函数：

[o, s, s2], updates = theano.scan(
            forward_prop_step,
            sequences=x,
            truncate_gradient=self.bptt_truncate,
            outputs_info=[None, 
                          dict(initial=T.zeros(self.hidden_dim)),
dict(initial=T.zeros(self.hidden_dim))])

tensorflow中也有一个

scan

函数，但它们得到的参数不同。如何将上面的

scan

函数转换为tensorflow

scan

函数