HeteGraph-Mamba: Heterogeneous Graph Learning via Selective State Space Model
This addresses the problem of learning from heterogeneous graphs (with different node types) for researchers and practitioners, representing a novel application rather than an incremental improvement.
The paper tackles heterogeneous graph learning by proposing HeteGraph-Mamba, the first method to apply selective state space models to this domain, which outperforms 19 state-of-the-art methods on benchmarks in both accuracy and efficiency.
We propose a heterogeneous graph mamba network (HGMN) as the first exploration in leveraging the selective state space models (SSSMs) for heterogeneous graph learning. Compared with the literature, our HGMN overcomes two major challenges: (i) capturing long-range dependencies among heterogeneous nodes and (ii) adapting SSSMs to heterogeneous graph data. Our key contribution is a general graph architecture that can solve heterogeneous nodes in real-world scenarios, followed an efficient flow. Methodologically, we introduce a two-level efficient tokenization approach that first captures long-range dependencies within identical node types, and subsequently across all node types. Empirically, we conduct comparisons between our framework and 19 state-of-the-art methods on the heterogeneous benchmarks. The extensive comparisons demonstrate that our framework outperforms other methods in both the accuracy and efficiency dimensions.