analiza aiqbench
This paper introduces HetSheaf, a novel framework that leverages cellular sheaves to model heterogeneity directly within the data structure of graphs rather than relying on specialized architectural modules. By incorporating type-aware local feature spaces and a universal readout mechanism, the approach significantly improves performance on node classification and link prediction tasks while reducing parameter count by up to tenfold compared to existing baselines. The work advances the theoretical foundations of graph neural networks by addressing complex real-world domains like biology and social networks with greater efficiency and accuracy.
Heterogeneous Sheaf Neural Networks
arXiv:2409.08036v4 Announce Type: replace Abstract: Heterogeneous graphs, whose nodes and edges can belong to different types and feature spaces, arise in many real-world domains, including biology, recommendation, social networks, and computer systems. Existing heterogeneous graph neural networks typically handle this heterogeneity at the architectural level through relation-specific modules, meta-path machinery or type-aware attention, which often leads to increasingly specialised parameter-heavy designs. In this work, we propose HetSheaf, a framework for learning heterogeneous graphs through cellular sheaves. Instead of encoding heterogeneity solely in the architecture, HetSheaf represents it directly in the underlying data structure by assigning type-aware local feature spaces and learning restriction maps conditioned on node features, node types, and edge types. To support graph-level prediction, we further introduce SheafPool, a universal stalk-space readout that aggregates node representations while being invariant to local changes of basis, thereby making graph classification with sheaf networks well-defined and achieving an F1 Score up to 42 percentage points higher than mean pooling. Across a diverse suite of benchmarks (node classification, link prediction and graph classification). HetSheaf consistently achieves up to 2 percentage points higher performance (up to 94.97% Macro F1 Score on node classification and up to 99.62% on link prediction) on the Heterogeneous Graph Benchmark (HGB) framework against homogeneous (GCN, GAT, GIN, GraphSAGE), heterogeneous (R-GCN, HAT, HGT) and type-agnostic sheaf baselines, while reducing the number of parameters by up to 10$\times$.
Abstract:Heterogeneous graphs, whose nodes and edges can belong to different types and feature spaces, arise in many real-world domains, including biology, recommendation, social networks, and computer systems. Existing heterogeneous graph neural networks typically handle this heterogeneity at the architectural level through relation-specific modules, meta-path machinery or type-aware attention, which often leads to increasingly specialised parameter-heavy designs. In this work, we propose HetSheaf, a framework for learning heterogeneous graphs through cellular sheaves. Instead of encoding heterogeneity solely in the architecture, HetSheaf represents it directly in the underlying data structure by assigning type-aware local feature spaces and learning restriction maps conditioned on node features, node types, and edge types. To support graph-level prediction, we further introduce SheafPool, a universal stalk-space readout that aggregates node representations while being invariant to local changes of basis, thereby making graph classification with sheaf networks well-defined and achieving an F1 Score up to 42 percentage points higher than mean pooling. Across a diverse suite of benchmarks (node classification, link prediction and graph classification). HetSheaf consistently achieves up to 2 percentage points higher performance (up to 94.97% Macro F1 Score on node classification and up to 99.62% on link prediction) on the Heterogeneous Graph Benchmark (HGB) framework against homogeneous (GCN, GAT, GIN, GraphSAGE), heterogeneous (R-GCN, HAT, HGT) and type-agnostic sheaf baselines, while reducing the number of parameters by up to 10$\times$.
Machine Learning (cs.LG)
arXiv:2409.08036 [cs.LG]
(or arXiv:2409.08036v4 [cs.LG] for this version)
From: Alessio Borgi Dr. [view email] [v1] Thu, 12 Sep 2024 13:38:08 UTC (596 KB) [v2] Thu, 16 Apr 2026 21:00:42 UTC (275 KB) [v3] Fri, 22 May 2026 09:52:38 UTC (372 KB) [v4] Mon, 19 Oct 2026 10:49:05 UTC (389 KB)
Text integral sub licența deschisă a publicației · 276 cuvinte
Subiecte conexe
Robust and efficient enriched representation with intrinsic neuronal heterogeneity
arXiv · 30 oct. 2026
Chi-Square Wavelet Graph Neural Networks for Heterogeneous Graph Anomaly Detection
arXiv · 23 iul. 2026
HiSE: A Lightweight Hierarchical Semantic Explainer for Heterogeneous Graph Neural Networks
arXiv · 3 iun. 2026
Virtual Smart Metering in District Heating Networks via Heterogeneous Spatial-Temporal Graph Neural Networks
arXiv · 10 oct. 2026