PACE: Propagation-Aware Collaborative Correction for One-Shot Personalized Federated Graph Learning

SummaryAI generated

On September 7, 2026, arXiv cs.LG published the paper PACE, which proposes a propagation-aware collaborative correction method for single-step personalized federated learning. This method aims to address the issues of limited information dissemination and asynchronous model updates in personalized federated learning. By introducing a propagation-awareness mechanism, it achieves collaborative correction between nodes, thereby improving the accuracy and convergence efficiency of personalized models.

Related eventsRELATED EVENTS

2026-09-09 05:29OpenAI claims huge maths breakthrough on a famed ‘Millennium Problem’
2026-09-09 06:11OpenAI claims to have solved the Navier-Stokes equation problem, but its originality is questioned by the academic community
2026-09-08 22:00Google DeepMind Maps 9 Billion Possible DNA Variants

Key entitiesKEY ENTITIES

PACE

SignalsSIGNALS

Keyword heat

PACE1

All reports (1)SOURCES

↗

A arXiv cs.LG en 2026-09-07 12:00

提出 PACE 方法，通过传播感知协同修正提升单轮个性化联邦图学习性能。该方法将协作知识视为对本地预测器的紧凑修正而非替代，客户端上传秩 -r 更新载体和对角线消息矩草图，服务器据此构建接收器锚定的修正项，接收器保留完整本地模型。利用凸负对数似然校准（CNLL）在验证节点上选择系数融合本地与外部 logits，参数固定且无需反馈。在六个数据集上，个性化返回占用 9.6-17.6% 稠密张量字节；在五个数据集上修正获得非零权重并提升准确率与加权 F1，仅在 ogbn-arxiv 上保留本地预测。