CoulGAT: An Experiment on Interpretability of Graph Attention Networks
This work addresses interpretability for researchers using graph neural networks, but it appears incremental as it builds on existing GAT methods with a new attention mechanism.
The authors tackled the problem of interpreting Graph Attention Networks (GAT) by developing CoulGAT, an attention mechanism inspired by the screened Coulomb potential, and applied it to train and analyze models on the CHAMPS dataset, resulting in a framework that extracts node-node and node-feature interactions to define an empirical standard model for graph structure and hidden layers.
We present an attention mechanism inspired from definition of screened Coulomb potential. This attention mechanism was used to interpret the Graph Attention (GAT) model layers and training dataset by using a flexible and scalable framework (CoulGAT) developed for this purpose. Using CoulGAT, a forest of plain and resnet models were trained and characterized using this attention mechanism against CHAMPS dataset. The learnable variables of the attention mechanism are used to extract node-node and node-feature interactions to define an empirical standard model for the graph structure and hidden layer. This representation of graph and hidden layers can be used as a tool to compare different models, optimize hidden layers and extract a compact definition of graph structure of the dataset.