Just Jump: Dynamic Neighborhood Aggregation in Graph Neural Networks

TL;DR

This paper introduces a dynamic neighborhood aggregation method guided by attention for graph neural networks, enabling selective, node-adaptive aggregation of neighboring embeddings to improve representation learning.

Contribution

It presents a novel DNA procedure with attention-guided, node-specific aggregation and channel-wise connection control to enhance GNN performance.

Findings

Effective in transductive node classification tasks

Outperforms existing GNN aggregation schemes

Reduces overfitting through grouped linear projections

Abstract

We propose a dynamic neighborhood aggregation (DNA) procedure guided by (multi-head) attention for representation learning on graphs. In contrast to current graph neural networks which follow a simple neighborhood aggregation scheme, our DNA procedure allows for a selective and node-adaptive aggregation of neighboring embeddings of potentially differing locality. In order to avoid overfitting, we propose to control the channel-wise connections between input and output by making use of grouped linear projections. In a number of transductive node-classification experiments, we demonstrate the effectiveness of our approach.