Updated docstring for shape info

[saiden89]

As proposed in #3573, here is my take on the first chunk of convolutional layers in PyG. More will follow soon.

[codecov-commenter]

Codecov Report

Merging #3697 (b542068) into master (ba468cd) will increase coverage by 0.00%.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##           master    #3697   +/-   ##
=======================================
  Coverage   81.68%   81.68%           
=======================================
  Files         295      295           
  Lines       14893    14898    +5     
=======================================
+ Hits        12165    12170    +5     
  Misses       2728     2728           

Impacted Files	Coverage Δ
torch_geometric/nn/conv/agnn_conv.py	100.00% <ø> (ø)
torch_geometric/nn/conv/appnp.py	78.33% <ø> (ø)
torch_geometric/nn/conv/arma_conv.py	100.00% <ø> (ø)
torch_geometric/nn/conv/cheb_conv.py	100.00% <ø> (ø)
torch_geometric/nn/conv/dna_conv.py	98.54% <ø> (ø)
torch_geometric/nn/conv/edge_conv.py	100.00% <ø> (ø)
torch_geometric/nn/conv/eg_conv.py	90.51% <ø> (ø)
torch_geometric/nn/conv/fa_conv.py	89.47% <ø> (ø)
torch_geometric/nn/conv/feast_conv.py	100.00% <ø> (ø)
torch_geometric/nn/conv/film_conv.py	96.92% <ø> (ø)
... and 5 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update ba468cd...b542068. Read the comment docs.

[saiden89]

Thanks for the corrections. Indeed, I was also wondering how to solve some inconsistencies in notation. So I guess now all features are F instead of H. How about the attention heads? Some papers refer to them as K (like GAT) some others to H. Do you think it would be better to follow the original article of to use a universal notation in PyG?

[rusty1s]

This is great, thank you :)
I further changed the CSS layout and made some style changes.

[rusty1s]

I think it's better to have a universal notation, and I like the following:

• |V|: number of nodes
• |E|: number of edges
• F: node feature size
• D: edge feature size
• H: number of heads
• L: number of layers

[saiden89]

I have another batch of the updated docstring. Is it possible to reopen this pull request or do I make a new one?

[rusty1s]

Please send a new one if possible :)