| 摘要: |
| 作为机器学习的一个新兴领域, 多源部分域适应(MSPDA)问题由于其源域自身的复杂性、领域之间的差异性以及目标域自身的无监督性, 给相关研究带来了挑战, 以致目前鲜有相关工作被提出. 在该场景下, 多个源域中的无关类样本在域适应过程中会造成较大的累积误差和负迁移. 此外, 现有多源域适应方法大多未考虑不同源域对目标域任务的贡献度不同. 因此, 提出基于自适应权重的多源部分域适应方法(AWMSPDA).首先, 构建了多样性特征提取器以有效利用源域的先验知识; 同时, 设计了多层次分布对齐策略从不同层面消除了分布差异,促进了正迁移; 此外, 为量化不同源域贡献度以及过滤源域无关类样本, 利用相似性度量以及伪标签加权方式构建自适应权重; 最后, 通过大量实验验证了所提出AWMSPDA算法的泛化性以及优越性. |
| 关键词: 多源部分域适应 负迁移 多样性特征提取 多层次分布对齐 自适应权重 |
| DOI：10.13328/j.cnki.jos.007015 |
| 分类号: |
| 基金项目:国家自然科学基金(62176128); 计算机软件新技术国家重点实验室开放课题(KFKT2022B06); 中央高校基本科研业务费(NJ2022028); 江苏省“青蓝工程”优秀青年骨干教师人才计划; 江苏省研究生科研实践创新计划(KYCX22_1205) |
|
| Adaptive Weight-induced Multi-source Partial Domain Adaptation |
|
TIAN Qing1,2,3, SUN Can-Yu1, CHU Yi1
|
|
1.School of Software, Nanjing University of Information Science Technology, Nanjing 210044, China;2.Engineering Research Center of Digital Forensics(Nanjing University of Information Science Technology), Ministry of Education, Nanjing 210044, China;3.State Key Laboratory for Novel Software Technology(Nanjing University), Nanjing 210023, China
|
| Abstract: |
| As an emerging field of machine learning, multi-source partial domain adaptation (MSPDA) poses challenges to related research due to the complexities of the involved source domains, the diversities between the domains, and the unsupervised nature of the target domain itself, leading to rarely few works presented. In this scenario, the irrelevant class samples in multiple-source domains will cause large cumulative errors and negative transfer during domain adaptation. In addition, most of the existing multisource domain adaptation methods do not consider the different contributions of different source domains to the target domain tasks. Therefore, this study proposes an adaptive weightinduced multisource partial domain adaptation (AW-MSPDA). Firstly, a diverse feature extractor is constructed to effectively utilize the prior knowledge of the source domain. Meanwhile, multilevel distribution alignment strategy is constructed to eliminate distribution discrepancies from different levels to promote positive transfer. Moreover, the pseudolabel weighting and similarity measurement are used to construct adaptive weights to quantify the contribution of different source domains and filter samples which are irrelevant to the source domain. Finally, the generalization and performance superiority of the proposed AWMSPDA algorithm are evaluated by extensive experiments. |
| Key words: multi-source partial domain adaptation negative transfer diverse feature extraction multilevel distribution alignment adaptive weight |
