1

I've been reading the deconvolution article on distill. I am not able to figure out the meaning of the text

These artifacts tend to be most prominent when outputting unusual colors. Since neural network layers typically have a bias (a learned value added to the output) it’s easy to output the average color. The further a color — like bright red — is away from the average color, the more deconvolution needs to contribute.

How would the deconvolution need to contribute more. And more specifically what does it mean to contribute more.

Thanks in advance

pramesh
  • 123
  • 5
  • I'd guess that *contributing more* means *outputting a higher intensity value* (like bright red as the author states). In that case, it makes sense that if the bias represent somewhat the mean color value of the output, then the operation of deconvolution should contribute more $\to$ Increasing the output value to get the desired bright color. – Javier TG Jun 12 '21 at 18:47

0 Answers0