GraphNN

Test Repo for Graph Transformer Models

TODO

Retest all modules so they actually work, most should at the moment

Purpose

This repo represents a set of torch modules used to manipulate graph information. Usually, we have a graph G, with a set of vertices V. Each vertex has a feature vector E, representing the embedding size. For example, using a 300 TF-IDF or BOW representation on a graph with 1440 vertices results in a torch tensor of size (G, V, E) -> (1, 1440, 300).

The idea here being, with an adjacency matrix G, V, V, where it is a binary tensor such that Row -> Col. With this, we can easily mask nodes in the multihead attention.

A sample model would be:

Nodes -> Linear Projection -> N Graph Attention Layers -> PairwiseBilinear

for a link prediction task, and use negative sampling to train.

For classification, an output dense layer should suffice,

Modules

• Graph Attention Layer
  • GraphAttentionNetwork
  • GraphAttentionLayer
• Node Transformer
  • PositionalEncoding
  • NodeTransformer
  • TransformerNetwork
  • These are unstable due to the large amount of memory used (1, V, S, E)
• Transformer
  • Transformer
  • TransformerDecoder
    • Still need to figure out use case for Networks
• Multihead Attention
  • MultiheadAttention
• Residual
  • AdditiveResidual
  • GeneralResidual
  • DotResidual
• Pairwise
  • PairwiseBilinear
  • PairwiseDot
  • PairwiseDistance