Incorrect behaviour of tf.nn.sigmoid on complex inputs

[jshepherd01]

Issue type

Bug

Have you reproduced the bug with TensorFlow Nightly?

Yes

Source

binary

TensorFlow version

tf v1.12.1-96406-gfa4d29bfef8 2.14.0-dev20230706

Custom code

Yes

OS platform and distribution

WSL Ubuntu 20.04.5 LTS

Mobile device

No response

Python version

No response

Bazel version

No response

GCC/compiler version

No response

CUDA/cuDNN version

No response

GPU model and memory

No response

Current behavior?

tf.nn.sigmoid gives incorrect values for complex numbers with large, negative real parts (which should map to within a rounding error of 0). For example, when x = -709-1j (complex128) I expect tf.nn.sigmoid(x) = 0, but instead tf.nn.sigmoid(x) = 1.

Standalone code to reproduce the issue

import tensorflow as tf
x = tf.constant([-709-1j], dtype=tf.complex128)
tf.nn.sigmoid(x)  # <tf.Tensor: shape=(1,), dtype=complex128, numpy=array([1.+0.j])>

Relevant log output

No response

[sushreebarsa]

@sachinprasadhs I was able to replicate the issue on colab using TF v2.13, tf-nightly. Please find the gist here for reference. Thank you!

[dlibk]

The same issue happens on tf.Tensor((-inf+0j), shape=(), dtype=complex128), the result is 1+0j instead of 0.
Please see the code below.

input = tf.constant(-np.inf, dtype='complex128')
print(input)  # tf.Tensor((-inf+0j), shape=(), dtype=complex128)

out = tf.nn.sigmoid(input)
print(out)  # tf.Tensor((1+0j), shape=(), dtype=complex128)

[LakshmiKalaKadali]

Hi @jshepherd01,

The tf.nn.sigmoid is working as intended for complex inputs. Checked with different examples in TF 2.17 and TF 2.18.0.

test_inputs = [
        tf.constant([-np.inf], dtype=tf.complex128),
        tf.constant([-709-1j], dtype=tf.complex128),
        tf.constant([-709+1j], dtype=tf.complex128),
        tf.constant([-1-1j], dtype=tf.complex128),
        tf.constant([-1+1j], dtype=tf.complex128),
        tf.constant([0-1j], dtype=tf.complex128),
        tf.constant([0+1j], dtype=tf.complex128),
        
    ]
    
print("Complex Sigmoid Investigations:")
for x in test_inputs:
  result = tf.nn.sigmoid(x)
  print(f"\nInput: {x.numpy()}")
  print(f"Sigmoid Output: {result.numpy()}")

OUTPUT

Sigmoid Output: [0.+0.j]

Input: [-709.-1.j]
Sigmoid Output: [0.-0.j]

Input: [-709.+1.j]
Sigmoid Output: [0.+0.j]

Input: [-1.-1.j]
Sigmoid Output: [0.21795843-0.20194823j]

Input: [-1.+1.j]
Sigmoid Output: [0.21795843+0.20194823j]

Input: [0.-1.j]
Sigmoid Output: [0.5-0.27315124j]

Input: [0.+1.j]
Sigmoid Output: [0.5+0.27315124j]

Please refer to the gist as well.

Thank You

[github-actions]

This issue is stale because it has been open for 7 days with no activity. It will be closed if no further activity occurs. Thank you.

[github-actions]

This issue was closed because it has been inactive for 7 days since being marked as stale. Please reopen if you'd like to work on this further.

[google-ml-butler]

Are you satisfied with the resolution of your issue?
Yes
No