Skip to content

Learnable epsilon in GIN - #697

Open
aurorarossi wants to merge 7 commits into
JuliaGraphs:masterfrom
aurorarossi:learnable-epsilon-gin
Open

aurorarossi wants to merge 7 commits into
JuliaGraphs:masterfrom
aurorarossi:learnable-epsilon-gin

Conversation

@aurorarossi

Copy link
Copy Markdown
Member

This PR adds PyTorch Geometric-style ϵ and train_eps options to GINConv for both Flux and Lux.

GINConv(nn; train_eps = true)  # trainable eps, initialized to 0
GINConv(nn;  ϵ  = 0.1f0)       # fixed eps

@CarloLucibello

Copy link
Copy Markdown
Member

needs a rebase

Comment thread GNNlib/src/layers/conv.jl
Comment thread GNNLux/src/layers/conv.jl Outdated
Comment thread GraphNeuralNetworks/src/layers/conv.jl Outdated
@aurorarossi
aurorarossi force-pushed the learnable-epsilon-gin branch 2 times, most recently from f0c8676 to 1d99de0 Compare September 30, 2026 13:09

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants