Meta Operator for Complex Query Answering on Knowledge Graphs
Hang Yin, Zihao Wang, Yangqiu Song
TL;DR
The paper tackles complex query answering over incomplete knowledge graphs under few-shot data by introducing Model Agnostic Meta Operator (MAMO), a meta-learning approach focused on operator-level generalization rather than query-type generalization. By learning a meta projection and adapting it to specific operator instances (e.g., θ_p, θ_i, θ_e) through inner adaptation and outer meta-optimization, MAMO achieves better combinatorial generalization across diverse query types. The authors provide a mathematical interpretation showing gradient decomposition across operator-type partitions, and demonstrate across multiple backbones (LogicE, FuzzQE, ConE) and datasets that MAMO consistently improves performance, especially in EPFO and EFO-1 settings. This operator-centric meta-learning framework offers a practical, model-agnostic path to robust CQA in data-scarce regimes with potential for broader applicability in KG reasoning tasks.
Abstract
Knowledge graphs contain informative factual knowledge but are considered incomplete. To answer complex queries under incomplete knowledge, learning-based Complex Query Answering (CQA) models are proposed to directly learn from the query-answer samples to avoid the direct traversal of incomplete graph data. Existing works formulate the training of complex query answering models as multi-task learning and require a large number of training samples. In this work, we explore the compositional structure of complex queries and argue that the different logical operator types, rather than the different complex query types, are the key to improving generalizability. Accordingly, we propose a meta-learning algorithm to learn the meta-operators with limited data and adapt them to different instances of operators under various complex queries. Empirical results show that learning meta-operators is more effective than learning original CQA or meta-CQA models.