Concatenate 3d tensors by cloning one tensor?

Question

I have two tensors:

a = tf.placeholder(tf.float32, [None, 20, 100])
b = tf.placeholder(tf.float32, [None, 1, 100])

I want to append b to a[i, 20, 100] , to create c such as c has a shape of [None, 20, 200] .

This seems rather simple but I haven't figured out how to do this with tf.concat :

tf.concat(0, [a, b]) -> Shapes (20, 100) and (1, 100) are not compatible
tf.concat(1, [a, b]) => shape=(?, 28, 100) which is not what I wanted
tf.concat(2, [a, b]) -> Shapes (?, 20) and (?, 1) are not compatible

Do I need to reshape a and b first then concat?

Answer 1

This can be done using tf.tile . You will need to clone the tensor along dimension 1, 20 times to make it compatible with a . Then a simple concatenation along dimension 2 will give you the result.

Here is the complete code,

import tensorflow as tf

a = tf.placeholder(tf.float32, [None, 20, 100])
b = tf.placeholder(tf.float32, [None, 1, 100])
c = tf.tile(b, [1,20,1])
print c.get_shape()
# Output - (?, 20, 100)
d = tf.concat(2, [a,c])
print d.get_shape()
# Output - (?, 20, 200)