四川轻化工大学学报（自然科学版）2025，Vol.38Issue(6)：39-50,12.DOI:10.11863/j.suse.2025.06.05
结合密度峰值聚类的主动学习研究综述
A Review of Active Learning Research Combined with Density Peak Clustering
摘要
Abstract
Machine learning is widely used,while traditional supervised learning relies on large amounts of labeled data.In reality,most data are label-free and manual tagging costs are high.Active learning can reduce tagging costs by selecting valuable samples.The active learning method combined with density peak clustering aims to optimize the sample selection process and improve the sample utilization and selection efficiency by using density peak clustering algorithm.Firstly,the basic contents of density peak clustering and active learning are introduced,five active learning methods of density peak clustering are enumerated and explained,which are active learning through density peak clustering,active learning through two-stage clustering,cooperative training algorithm combining active learning and density peak clustering,active learning based on density clustering and neighborhood,and boundary sampling with density entropy.Then,the applications of the active learning method combined with density peak clustering in image segmentation and classification,data stream classification and network attack attribution are discussed.Finally,the research hotspots and the future development direction are prospected.关键词
主动学习/密度峰值聚类/标注样本Key words
active learning/density peak clustering/sample labeling分类
信息技术与安全科学引用本文复制引用
李茜苒,王继奎..结合密度峰值聚类的主动学习研究综述[J].四川轻化工大学学报（自然科学版）,2025,38(6):39-50,12.基金项目
国家自然科学基金项目(12201267) （12201267）
甘肃省自然科学基金项目(24JRRA1169) （24JRRA1169）
