Tensorflow中的过滤张量元素

Question

What's the equivalent operation in Tensorflow for this? For example, I have a x = np.array([-12,4,6,8,100]) . I want to do as simple as this: x = x[x>5] , but I can't find any TF operation for this. Thanks!

Answer 1

In TF you can do something like this to achieve similar results.

import numpy as np 
import tensorflow as tf 

x = np.array([-12,4,6,8,100])
y = tf.gather(x, tf.where(x > 5))
y.numpy().reshape(-1)
array([  6,   8, 100])

Details

The tf.where will return the indices of condition that are True . Such as

x = np.array([-12,4,6,8,100])
tf.where(x > 5)

<tf.Tensor: shape=(3, 1), dtype=int64, numpy=
array([[2],
       [3],
       [4]])>

And next, using tf.gather , it slices from params ( x ) axis according to indices (from tf.where ). Such as

tf.gather(x, tf.where(x > 5))