arXiv 2605.105102026-05-11cs.LGcs.AI

CMKL：面向演变生物医学知识图谱的模态感知持续学习

CMKL: Modality-Aware Continual Learning for Evolving Biomedical Knowledge Graphs

Technology, Innovation, Entrepreneurship Department（技术、创新与创业部门）；King Abdullah University of Science and Technology（卡塔尔科技大学）；School of Computing and Information Systems（计算与信息系统学院）；The University of Melbourne（墨尔本大学）；College of Computer Science and Technology（计算机科学与技术学院）；Jilin University（吉林大学）；School of Computing Technologies（计算技术学院）；RMIT University（皇家墨尔本理工大学）；College of Computing and Data Science（计算与数据科学学院）；Nanyang Technological University（南洋理工大学）

Yousef A. Radwan, Yao Li, Qing Qing, Ziqi Xu, Qixin Zhang, Yongcheng Jing, Renqiang Luo, Xikun Zhang

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

CMKL通过混合专家路由融合结构、文本和分子信息，利用EWC和K-means多样化重放缓冲区保护已有知识，在生物医学持续学习任务中取得显著提升。

摘要：

生物医学知识图谱日益庞大、动态且多模态，受高通量测序等生物技术进步推动。机器学习模型可推断未观测到的生物医学关系并表征图中生物实体，但现有知识图谱嵌入方法及其持续学习扩展要么假设静态图结构，要么在演变数据分布下无法利用多模态信息。它们还对所有模型参数应用统一正则化，忽视不同模态在图演变中可能表现出不同的遗忘动态。我们提出了持续多模态知识图谱学习器（CMKL），一种针对生物医学KG的CL框架，原生编码结构、文本和分子，通过混合专家（MoE）路由融合它们，并通过标准EWC正则化和K-means多样化多模态重放缓冲区保护已有知识。我们在包含129K实体的生物医学持续学习基准上评估CMKL，共10个任务。在持续生物医学实体分类中，CMKL达到AP 0.591，优于最强的结构基线0.370，提升60%，由多模态特征访问和序列中接近零遗忘（AF 0.008）驱动。在持续关系预测中，CMKL达到AP $0.062$，在种子噪声下匹配Naive Sequential和EWC（0.058），优于Joint Training（0.047, p=0.045）和LKGE（0.039）。冻结文本消融达到AP 0.136，超过任何联合训练模型两倍以上，但该信号无法通过边际排名梯度达到：贪婪模态不对称性存在于表示层面，而非融合层面，MoE路由通过抑制不可达模态而不强制其通过学习瓶颈来管理它。代码：github.com/yradwan147/cmkl-neurips2026

英文摘要：

Biomedical knowledge graphs are increasingly large, dynamic, and multimodal, driven by rapid advances in biotechnology such as high-throughput sequencing. Machine learning models can infer previously unobserved biomedical relationships and characterize biomedical entities in these graphs, but existing knowledge graph embedding methods and their continual learning extensions either assume static graph structure or fail to exploit multimodal information under evolving data distributions. They also apply uniform regularization across all model parameters, ignoring that different modalities may exhibit distinct forgetting dynamics as the graph evolves. We propose the Continual Multimodal Knowledge Graph Learner (CMKL), a CL framework for biomedical KGs that natively encodes structure, text, and molecules, fuses them through a Mixture-of-Experts (MoE) router, and protects previously learned knowledge with standard EWC regularization and a K-means-diverse multimodal replay buffer. We evaluate CMKL on a 129K-entity biomedical continual benchmark with 10 tasks. On continual biomedical entity classification, CMKL reaches AP 0.591 versus 0.370 for the strongest structural baseline, a 60% gain that is driven by access to multimodal features and preserved across the sequence with near-zero forgetting (AF 0.008). On continual relationship prediction, CMKL reaches AP $0.062$, matching Naive Sequential and EWC (0.058) within seed noise and outperforming Joint Training (0.047, p=0.045) and LKGE (0.039). A frozen-text ablation reaches AP 0.136, more than double any jointly trained model, yet that signal is unreachable by margin-ranking gradients: the greedy-modality asymmetry lives at the representation level, not the fusion level, and MoE routing manages it by suppressing the unreachable modality without forcing it through a learned bottleneck. Code: github.com/yradwan147/cmkl-neurips2026

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。