Closed kristian-georgiev closed 3 years ago
Hi @kristian-georgiev
Thanks for your message! This is indeed an artifact of the pooling and dilation. Using SO(2) equivariance is not going to solve this, unfortunately.
Have a look at page 3 here: https://arxiv.org/pdf/2004.09691.pdf We tried to explain a bit this issue and how to solve that there. Let me know if this helps
Best, Gabriele
This makes perfect sense, thank you!
Hi Gabriele, thank you very much for the great library!
Is equivariance guaranteed for images of arbitrary sizes/resolutions? It seems like equivariance is satisfied only for certain sizes. As an example, the
C4
-invariant10-4 Wide ResNet
defined and tested inexamples/e2wrn.py
seems to only be invariant to90
-degree rotations for images with widths of the form4k+3
.To exemplify this, the output of this snippet
is
This seems like an artifact caused by dilation/pooling but I am not sure whether there's any other factor causing this behaviour. Is this the case?
Lastly, would we observe a similar behaviour if we are working with field types composed of irreps of
SO(2)
? I don't have an intuition on how dilation is going to interact with guaranteeing equivariance there.Thanks in advance!