🤖 AI Summary
This work addresses two fundamental limitations of Message Passing Neural Networks (MPNNs): oversquashing and poor modeling of long-range interactions. We propose a topology-aware, sensitivity-adaptive virtual node (VN) method. Unlike conventional VNs that assign uniform weights to all nodes, our approach leverages graph mixing dynamics and sensitivity theory to design a structural-aware, dynamic weighting mechanism—retaining the original computational complexity while overcoming the constraint of weight homogenization. Theoretically, we show that the VN alleviates oversquashing by enhancing graph mixing capability. Empirically, extensive evaluations across multiple graph benchmarks demonstrate that our novel VN variant achieves significant performance gains on graph-level tasks with minimal overhead, establishing itself as an efficient and strong baseline. It consistently outperforms classical VN baselines and several Graph Transformer variants.
📝 Abstract
While message passing neural networks (MPNNs) have convincing success in a range of applications, they exhibit limitations such as the oversquashing problem and their inability to capture long-range interactions. Augmenting MPNNs with a virtual node (VN) removes the locality constraint of the layer aggregation and has been found to improve performance on a range of benchmarks. We provide a comprehensive theoretical analysis of the role of VNs and benefits thereof, through the lenses of oversquashing and sensitivity analysis. First, we characterize, precisely, how the improvement afforded by VNs on the mixing abilities of the network and hence in mitigating oversquashing, depends on the underlying topology. We then highlight that, unlike Graph-Transformers (GTs), classical instantiations of the VN are often constrained to assign uniform importance to different nodes. Consequently, we propose a variant of VN with the same computational complexity, which can have different sensitivity to nodes based on the graph structure. We show that this is an extremely effective and computationally efficient baseline for graph-level tasks.