FedEAN: Entity-Aware Adversarial Negative Sampling for Federated Knowledge Graph Reasoning

Federated knowledge graph reasoning (FedKGR) aims to perform reasoning over different clients while protecting data privacy, drawing increasing attention to its high practical value. Previous works primarily focus on data heterogeneity, ignoring challenges from limited data scale and primitive negative sample strategies, i.e., random entity replacement, which yield low-quality negatives and zero loss issues. Meanwhile, generative adversarial networks (GANs) are widely used in different fields to generate high-quality negative samples, but no work has been developed for FedKGR. To this end, we propose a plug-and-play Entity-aware Adversarial Negative sampling strategy for FedKGR, termed FedEAN. Specifically, we are the first to adopt GANs to generate high-quality negative samples in different clients. It takes the target triplet in each batch as input and outputs high-quality negative samples, which guaranteed by the joint training of the generator and discriminator. Moreover, we design an entity-aware adaptive negative sampling mechanism based on the similarity of entity representations before and after server aggregation, which can persevere the entity global consistency across clients during training. Extensive experiments demonstrate that FedEAN excels with various FedKGR backbones, demonstrating its ability to construct high-quality negative samples and address the zero-loss issue.

I. Introduction

Knowledge graphs (KGs) represent real-world knowledge in a structured manner [1], [2], [3], [4]. However, dynamic attributes and ongoing changes in KGs result in issues of fact incompleteness [5], [6], [7], [8], [9]. To address this issue, knowledge graph reasoning (KGR) has become a well-explored research area [10], [11], [12], [13], which aims to complete missing facts of KGs based on observed facts [7], [14], [15], [16]. However, most existing models access triplets from only a single device. In certain real world, KGs are more commonly decentralized in their existence considering data privacy and business interests, i.e., data is distributed and privatized across different devices, which is different from centralized learning where all KG data is accessible on a single device. To address this issue, Federated Learning (FL) has been introduced [17], [18]. This innovative machine learning approach involves multiple devices working together to train a global model using data from their local samples. As of now, federated learning has been widely experimented with on different types of data, including image [19], text [20] and graph [21]. However, the exploration in the domain of knowledge graphs is still insufficient. Take an example in Fig. 1, we can collectively and securely acquire the overall image of Sergy Brin holistically with the FL techniques, which significant accomplishments span computer science, business, and film, catalogued in distributed KGs such as Google knowledge base, IMDB, and Stanford database. For increasingly privatized knowledge graph data in the real world like Fig. 1, federated learning is an effective learning paradigm.