Mean Euclidean distance in Tensorflow

Question

I have two tensors of sequences of size [batch_size, seq_length, 2]. I want to compute mean Euclidean distance between tensors. What is the elegant way to do this?

Answer 1

You can also use tf.math.reduce_euclidean_norm:

tf.math.reduce_euclidean_norm(
    input_tensor, axis=None, keepdims=False, name=None
)

see the documentation here.

Answer 2

Given the two tensors A & B each with shape [batch_size, seq_length, 2], you can compute the Euclidean distance (L2 norm) using tf.norm:

l2_norm = tf.norm(A-B, ord='euclidean')