Rishabh

[WIP] Adaptive activation functions

6

Work-in-progress for [#355](https://github.com/SciML/NeuralPDE.jl/issues/355)

Added EdgeWeightNorm layer

1

Verified outputs with the pytorch implementation, will add tests now.

implementation of the DropMessage paper

2

Recently I had planned to work on some experiments related to the [DropMessage paper](https://arxiv.org/abs/2204.10037) and its extensions and application on various datasets like - Cora - CiteSeer - Pubmed -...

Add `drop_edge` transform

2

This is the second step in successfully implementing Issue https://github.com/CarloLucibello/GraphNeuralNetworks.jl/issues/412 for which we will need to implement [DropEdge](https://arxiv.org/abs/1907.10903). This PR can be merged after PR #414 is merged.

Added Personalized PageRank Diffusion [`ppr_diffusion` function]

3

Continues Issue #412, takes inspiration from [DGL implementation](https://docs.dgl.ai/en/0.8.x/generated/dgl.transforms.PPR.html) This PR introduces a new function, `ppr_diffusion`. It calculates Personalized PageRank (PPR) diffusion based on the edge weight matrix and updates the...

Added `perturb_edges` function

2

Relates to Issue #412 This is the first in the list of data augmentation based functionality we can add to ensure users can augment their data in many different ways...

Implements [Topology Adaptive Graph Convolutional Networks](https://arxiv.org/pdf/1710.10370.pdf) Computes: $$ H^{K} = \sum_{k=0}^K (D^{-1/2} A D^{-1/2})^{k} X \Theta_{k} $$ let me know if there are changes required in the method of computation

Next step to complete Issue #412 Will be ready for merge once PR #424 is merged Implements [DropNode](https://arxiv.org/pdf/2008.12578.pdf).

Adding Usage Examples for Graph Convolutional Layers

2

Currently, only a few convolutional layers have examples demonstrating how to use them, such as `GCNConv`, `CGConv`, `MEGNetConv`, `GNNGraph`, `SGConv`, and `EGNNConv` (6 out of 18 layers). We should aim...

Add AGNNConv support for HeteroGraphConv

3

Work in progress, will add tests soon. Please let me know if changes need to be made in the core code. Covers [issue #311](https://github.com/CarloLucibello/GraphNeuralNetworks.jl/issues/311)