What is the most efficient way to get the upper triangle of a symmetric matrix in Tensorflow/Keras?

Question

I have an intermediate model that outputs symmetric arrays. Those arrays are then used as input to another model. I'd like to just flatten the arrays and discard the lower triangles, since they're symmetric. Is there a best/most efficient way to do this?

Edit: I want the triangle extraction to be handled similar to any other Keras layer, so that the output of the first model can be input directly to the second model and trained end-to-end.

Possible duplicate of [extract upper/lower triangular part of a numpy matrix?](https://stackoverflow.com/questions/8905501/extract-upper-lower-triangular-part-of-a-numpy-matrix) — Joe, Feb 27 '18 at 08:12

score 2 · Accepted Answer · edited May 16 '19 at 11:32

2

Tensorflow and Keras use Numpy to handle the data. Consider using Numpy functions triu or triu_indices

edited May 16 '19 at 11:32

peter.babic

3,214
3
18
31

answered Feb 27 '18 at 08:10

Joe

6,758
2
26
47

Thanks. I wasn't able to use the NumPy functions directly, but the triu source code led me to an efficient implementation with a Lambda layer. – Jeremiah Johnson Feb 27 '18 at 18:59
@JeremiahJohnson Can you share your solution? Thanks – smörkex Nov 27 '21 at 18:21

What is the most efficient way to get the upper triangle of a symmetric matrix in Tensorflow/Keras?

1 Answers1