Quantum Graph Attention Networks: Trainable Quantum Encoders for Inductive Graph Learning

2025-09-14

AI Summary

To address the limited inductive generalization capability of quantum graph neural networks (QGNNs) in chemical property prediction, this work proposes Quantum Graph Attention Networks (QGATs)—a trainable quantum encoder. QGATs integrate parameterized quantum circuits with a quantum attention mechanism: unitary transformations dynamically weight neighborhood contributions, enabling joint quantum encoding of node features and topological structure; quantum state embedding and a quantum-classical hybrid training framework support end-to-end quantum learning on molecular graphs. On the QM9 benchmark, QGATs significantly outperform non-attention quantum baselines on larger molecular graphs and match classical Graph Attention Networks (GATs) on small molecules. These results demonstrate QGATs’ effectiveness as an efficient, scalable quantum encoder, with superior generalization and representational capacity for quantum graph learning.

Technology Category

• Machine Learning: Quantum Machine Learning
• Knowledge Representation and Reasoning: Qualitative Reasoning
• Computer Vision: Generative Adversarial Networks (GANs) for Vision

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

We introduce Quantum Graph Attention Networks (QGATs) as trainable quantum encoders for inductive learning on graphs, extending the Quantum Graph Neural Networks (QGNN) framework. QGATs leverage parameterized quantum circuits to encode node features and neighborhood structures, with quantum attention mechanisms modulating the contribution of each neighbor via dynamically learned unitaries. This allows for expressive, locality-aware quantum representations that can generalize across unseen graph instances. We evaluate our approach on the QM9 dataset, targeting the prediction of various chemical properties. Our experiments compare classical and quantum graph neural networks-with and without attention layers-demonstrating that attention consistently improves performance in both paradigms. Notably, we observe that quantum attention yields increasing benefits as graph size grows, with QGATs significantly outperforming their non-attentive quantum counterparts on larger molecular graphs. Furthermore, for smaller graphs, QGATs achieve predictive accuracy comparable to classical GAT models, highlighting their viability as expressive quantum encoders. These results show the potential of quantum attention mechanisms to enhance the inductive capacity of QGNN in chemistry and beyond.

Problem

• Extending quantum graph neural networks with trainable encoders
• Leveraging quantum attention for dynamic neighbor contribution modulation
• Enhancing inductive graph learning for molecular property prediction

Innovation

• Quantum attention mechanisms modulate neighbor contributions
• Parameterized quantum circuits encode node features
• Dynamic unitaries enable locality-aware quantum representations