NGDB-Zoo: Towards Efficient and Scalable Neural Graph Databases Training

2026-02-25

AI Summary

Existing neural graph databases suffer from low training efficiency and limited expressiveness due to query-level batching and structure-specific embeddings. This work proposes an operator-level dynamic scheduling training framework that decouples logical operators from query topology, enabling efficient integration of high-dimensional semantic priors from pretrained text encoders through dynamic dataflow scheduling and multi-stream parallel computation. By circumventing I/O bottlenecks and memory overflow, the approach achieves 1.8–6.8× throughput gains across six benchmarks, substantially improves GPU utilization, and effectively alleviates representational friction in hybrid neuro-symbolic reasoning.

Technology Category

• Natural Language Processing: Sentence-level Semantics, Textual Inference, etc.
• Planning, Routing, and Scheduling: Planning with Language Models
• Knowledge Representation and Reasoning: Computational Complexity of Reasoning

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

Neural Graph Databases (NGDBs) facilitate complex logical reasoning over incomplete knowledge structures, yet their training efficiency and expressivity are constrained by rigid query-level batching and structure-exclusive embeddings. We present NGDB-Zoo, a unified framework that resolves these bottlenecks by synergizing operator-level training with semantic augmentation. By decoupling logical operators from query topologies, NGDB-Zoo transforms the training loop into a dynamically scheduled data-flow execution, enabling multi-stream parallelism and achieving a $1.8\times$ - $6.8\times$ throughput compared to baselines. Furthermore, we formalize a decoupled architecture to integrate high-dimensional semantic priors from Pre-trained Text Encoders (PTEs) without triggering I/O stalls or memory overflows. Extensive evaluations on six benchmarks, including massive graphs like ogbl-wikikg2 and ATLAS-Wiki, demonstrate that NGDB-Zoo maintains high GPU utilization across diverse logical patterns and significantly mitigates representation friction in hybrid neuro-symbolic reasoning.

Problem

• Neural Graph Databases
• training efficiency
• expressivity
• query-level batching
• structure-exclusive embeddings

Innovation

• operator-level training
• semantic augmentation
• decoupled architecture
• multi-stream parallelism
• neural graph databases