《计算机应用》唯一官方网站 ›› 2025, Vol. 45 ›› Issue (9): 2727-2736.DOI: 10.11772/j.issn.1001-9081.2024091277
• 人工智能 •
收稿日期:
2024-09-09
修回日期:
2025-02-25
接受日期:
2025-03-03
发布日期:
2025-03-26
出版日期:
2025-09-10
通讯作者:
俞璐
作者简介:
王闯(1995—),男,安徽滁州人,硕士研究生,主要研究方向:迁移学习、模式识别基金资助:
Chuang WANG, Lu YU(), Jianwei CHEN, Cheng PAN, Wenbo DU
Received:
2024-09-09
Revised:
2025-02-25
Accepted:
2025-03-03
Online:
2025-03-26
Published:
2025-09-10
Contact:
Lu YU
About author:
WANG Chuang, born in 1995, M. S. candidate. His research interests include transfer learning, pattern recognition.Supported by:
摘要:
作为迁移学习的关键技术,域适应能很好地解决训练和测试数据集分布不同的问题。然而,传统的域适应方法通常只适用于目标域和源域数据集所含类别的数量和种类相同的情况,在实际场景中该条件通常很难满足。开集域适应(OSDA)正是为了解决此问题而出现的。为了填补该领域的空白,并为相关研究提供借鉴参考,对近年来出现的OSDA方法进行归纳分析。首先,介绍相关概念与基本结构;其次,分别从针对数据增强、针对特征提取以及针对分类器3个阶段梳理分析相关的典型方法;最后,对OSDA的未来发展方向进行展望。
中图分类号:
王闯, 俞璐, 陈健威, 潘成, 杜文博. 开集域适应综述[J]. 计算机应用, 2025, 45(9): 2727-2736.
Chuang WANG, Lu YU, Jianwei CHEN, Cheng PAN, Wenbo DU. Review of open set domain adaptation[J]. Journal of Computer Applications, 2025, 45(9): 2727-2736.
方法 | 基本原理 | 优点 | 局限性 | |
---|---|---|---|---|
针对数据 增强的方法 | 简单数据 增强方法 | 通过找寻源域已知类和目标域未知类数据之间的潜在关系,将源域不同已知类数据的不同组分按照一定方式合成代表未知类的数据,并用于后续模型训练 | 实现相对简单,可解释性较强,通常不会引入额外的神经网络结构,也不会增加模型复杂度和算力成本 | 需要源域已知类和目标域未知类数据特征的先验知识,否则很难找到潜在关系,导致合成的数据无法真正代表未知类,由此引发负迁移 |
神经网络 数据增强方法 | 通常需要引入GAN等额外的神经网络结构,通过设计具有针对性的策略,引导模型自动生成代表未知类的数据,并用于后续模型训练 | 神经网络自动学习源域已知类和目标域未知类数据的特征关系,即使对二者先验知识较少,也能实现 | 引入了额外神经网络,模型复杂度和计算成本增加;生成的未知类数据具有很强随机性,难以确保能很好代表未知类数据,从而导致负迁移 | |
针对特征 提取的方法 | 基于领域分布 差异的方法 | 通常需要通过模型训练先对目标域数据标记伪标签,而后通过减小源域和目标域已知类数据分布差异,同时增大和目标域未知类数据分布差异的方法减小目标域泛化误差 | 该类方法较为直观,可解释性较强,对不同数据集都有一定的兼容性 | 对伪标签质量的依赖较强,若伪标签质量较差,则容易产生负迁移;对域偏移较大的数据集效果较差,容易混淆未知类和域偏移部分 |
基于对抗学习的 方法 | 通过设定固定阈值或其他策略,对未知类数据进行鉴别,同时引入对抗训练思想,引导特征提取器学习源域数据和目标域已知类数据共有特征实现已知类对齐 | 能够有效降低源域和目标域已知类数据分布差异,通过对抗训练,能够使模型学习到更加鲁棒的特征表示,以提高模型在目标域上的泛化能力 | 模型复杂度和计算成本较高;多数基于对抗学习方法都设置了固定阈值,对参数选择较为敏感,不同开放程度的数据集阈值通常不同,需要进行多次实验进行调整选取 | |
基于语义分析的 方法 | 通常利用不同的聚类算法,提取源域和目标域不同类别数据所蕴含的语义信息,并用于指导模型训练 | 该类方法综合考虑了源域和目标域数据的语义信息,并以此进一步指导模型特征选择,能够较好提升模型性能 | 参数过多,计算量过大,计算效率较低;同时,该类方法对数据需求量较大,且对光照和噪声等因素较为敏感 | |
针对分类器的方法 | 基于对抗学习的 方法 | 通常先将目标域数据全部当成已知类,通过MCD方法进行特征提取器和分类器的交替训练,使分类边界更加清晰,而后通过额外设计未知类鉴别策略,对已经分为已知类的数据进行未知类鉴别 | 该类方法有利于模型进一步优化分类器的决策边界,减少目标域中边界样本数量,从而提高模型鲁棒性 | 模型复杂度和计算成本大幅增加,同时,由于目标域未知类数据的影响,使得分类器难以实现完全对齐 |
基于一对多 网络的方法 | 使用多个OVA网络代替多类分类器,模型训练后,目标域已知类数据将被分类为对应类别,目标域未知类数据将被分类为多个类别,或者不分类为任一类别 | 相对于多类分类器,OVA分类器自身具有识别未知类的能力,能够很好适应OSDA场景 | 多个OVA分类器,大幅提升了模型复杂度;在处理存在较大域偏移数据集时,性能难以保证 |
表1 OSDA方法的基本原理、优点和局限性
Tab. 1 Basic principles, advantages, and limitations of OSDA methods
方法 | 基本原理 | 优点 | 局限性 | |
---|---|---|---|---|
针对数据 增强的方法 | 简单数据 增强方法 | 通过找寻源域已知类和目标域未知类数据之间的潜在关系,将源域不同已知类数据的不同组分按照一定方式合成代表未知类的数据,并用于后续模型训练 | 实现相对简单,可解释性较强,通常不会引入额外的神经网络结构,也不会增加模型复杂度和算力成本 | 需要源域已知类和目标域未知类数据特征的先验知识,否则很难找到潜在关系,导致合成的数据无法真正代表未知类,由此引发负迁移 |
神经网络 数据增强方法 | 通常需要引入GAN等额外的神经网络结构,通过设计具有针对性的策略,引导模型自动生成代表未知类的数据,并用于后续模型训练 | 神经网络自动学习源域已知类和目标域未知类数据的特征关系,即使对二者先验知识较少,也能实现 | 引入了额外神经网络,模型复杂度和计算成本增加;生成的未知类数据具有很强随机性,难以确保能很好代表未知类数据,从而导致负迁移 | |
针对特征 提取的方法 | 基于领域分布 差异的方法 | 通常需要通过模型训练先对目标域数据标记伪标签,而后通过减小源域和目标域已知类数据分布差异,同时增大和目标域未知类数据分布差异的方法减小目标域泛化误差 | 该类方法较为直观,可解释性较强,对不同数据集都有一定的兼容性 | 对伪标签质量的依赖较强,若伪标签质量较差,则容易产生负迁移;对域偏移较大的数据集效果较差,容易混淆未知类和域偏移部分 |
基于对抗学习的 方法 | 通过设定固定阈值或其他策略,对未知类数据进行鉴别,同时引入对抗训练思想,引导特征提取器学习源域数据和目标域已知类数据共有特征实现已知类对齐 | 能够有效降低源域和目标域已知类数据分布差异,通过对抗训练,能够使模型学习到更加鲁棒的特征表示,以提高模型在目标域上的泛化能力 | 模型复杂度和计算成本较高;多数基于对抗学习方法都设置了固定阈值,对参数选择较为敏感,不同开放程度的数据集阈值通常不同,需要进行多次实验进行调整选取 | |
基于语义分析的 方法 | 通常利用不同的聚类算法,提取源域和目标域不同类别数据所蕴含的语义信息,并用于指导模型训练 | 该类方法综合考虑了源域和目标域数据的语义信息,并以此进一步指导模型特征选择,能够较好提升模型性能 | 参数过多,计算量过大,计算效率较低;同时,该类方法对数据需求量较大,且对光照和噪声等因素较为敏感 | |
针对分类器的方法 | 基于对抗学习的 方法 | 通常先将目标域数据全部当成已知类,通过MCD方法进行特征提取器和分类器的交替训练,使分类边界更加清晰,而后通过额外设计未知类鉴别策略,对已经分为已知类的数据进行未知类鉴别 | 该类方法有利于模型进一步优化分类器的决策边界,减少目标域中边界样本数量,从而提高模型鲁棒性 | 模型复杂度和计算成本大幅增加,同时,由于目标域未知类数据的影响,使得分类器难以实现完全对齐 |
基于一对多 网络的方法 | 使用多个OVA网络代替多类分类器,模型训练后,目标域已知类数据将被分类为对应类别,目标域未知类数据将被分类为多个类别,或者不分类为任一类别 | 相对于多类分类器,OVA分类器自身具有识别未知类的能力,能够很好适应OSDA场景 | 多个OVA分类器,大幅提升了模型复杂度;在处理存在较大域偏移数据集时,性能难以保证 |
方法 | 不同已知类的分类准确率 | UNK | OS* | HOS | |||||
---|---|---|---|---|---|---|---|---|---|
aeroplane | bicycle | bus | car | horse | knife | ||||
DANN[ | 93.7 | 80.4 | 89.7 | 71.0 | 92.5 | 65.2 | 0.0 | 82.1 | 0.0 |
AMS[ | 73.1 | 55.9 | 67.1 | 60.8 | 76.5 | 3.2 | 64.4 | 56.1 | 59.9 |
DAMC[ | 48.2 | 38.4 | 15.4 | 28.6 | 46.6 | 9.9 | 72.1 | 31.2 | 43.5 |
DCC[ | 75.8 | 57.7 | 83.9 | 68.7 | 76.3 | 83.5 | 32.0 | 74.3 | 44.8 |
文献[ | 86.9 | 58.6 | 77.2 | 63.3 | 86.0 | 8.8 | 52.9 | 63.5 | 57.7 |
文献[ | 56.0 | 56.1 | 69.2 | 57.8 | 66.6 | 28.8 | 73.9 | 55.7 | 63.6 |
表2 在VisDA数据集上各OSDA方法的结果 (%)
Tab. 2 Results of various OSDA methods on VisDA dataset
方法 | 不同已知类的分类准确率 | UNK | OS* | HOS | |||||
---|---|---|---|---|---|---|---|---|---|
aeroplane | bicycle | bus | car | horse | knife | ||||
DANN[ | 93.7 | 80.4 | 89.7 | 71.0 | 92.5 | 65.2 | 0.0 | 82.1 | 0.0 |
AMS[ | 73.1 | 55.9 | 67.1 | 60.8 | 76.5 | 3.2 | 64.4 | 56.1 | 59.9 |
DAMC[ | 48.2 | 38.4 | 15.4 | 28.6 | 46.6 | 9.9 | 72.1 | 31.2 | 43.5 |
DCC[ | 75.8 | 57.7 | 83.9 | 68.7 | 76.3 | 83.5 | 32.0 | 74.3 | 44.8 |
文献[ | 86.9 | 58.6 | 77.2 | 63.3 | 86.0 | 8.8 | 52.9 | 63.5 | 57.7 |
文献[ | 56.0 | 56.1 | 69.2 | 57.8 | 66.6 | 28.8 | 73.9 | 55.7 | 63.6 |
方法 | 是否端到端训练 | A-D | A-W | D-A | ||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
OS* | UNK | HOS | OS* | UNK | HOS | OS* | UNK | HOS | ||||
DANN[ | √ | 94.0 | — | — | 93.5 | — | — | 84.6 | — | — | ||
AMS[ | √ | 89.3 | 71.8 | 79.6 | 87.3 | 67.3 | 76.0 | 74.9 | 65.4 | 69.8 | ||
DAMC[ | √ | 19.1 | 84.0 | 31.1 | 25.2 | 88.9 | 39.3 | 21.9 | 92.3 | 35.4 | ||
DCC[ | √ | 91.3 | 41.3 | 56.9 | 92.3 | 43.6 | 59.2 | 76.2 | 70.0 | 73.0 | ||
文献[ | × | 89.1 | 52.6 | 66.2 | 85.7 | 43.8 | 57.9 | 76.3 | 52.8 | 62.4 | ||
文献[ | √ | 88.4 | 76.2 | 81.8 | 81.6 | 82.0 | 81.8 | 58.2 | 92.6 | 71.5 | ||
方法 | D-W | W-A | W-D | Avg | ||||||||
OS* | UNK | HOS | OS* | UNK | HOS | OS* | UNK | HOS | OS* | UNK | HOS | |
DANN[ | 100.0 | — | — | 78.1 | — | — | 100.0 | — | — | 91.7 | — | — |
AMS[ | 99.3 | 78.9 | 88.0 | 76.8 | 70.2 | 73.3 | 100.0 | 82.8 | 90.6 | 88.0 | 72.8 | 79.6 |
DAMC[ | 65.0 | 94.1 | 76.9 | 25.1 | 88.6 | 39.1 | 75.3 | 91.9 | 82.8 | 38.6 | 90.0 | 50.8 |
DCC[ | 100.0 | 66.0 | 79.5 | 72.5 | 60.8 | 66.2 | 100.0 | 66.0 | 79.5 | 88.7 | 59.5 | 70.1 |
文献[ | 100.0 | 34.8 | 51.7 | 77.0 | 49.1 | 60.0 | 100.0 | 46.2 | 63.2 | 88.0 | 46.6 | 60.2 |
文献[ | 99.2 | 93.4 | 96.2 | 60.6 | 92.5 | 73.2 | 100.0 | 93.0 | 96.4 | 81.3 | 88.3 | 83.5 |
表3 在Office-31数据集上各OSDA方法的结果 (%)
Tab. 3 Results of various OSDA methods on Office-31 dataset
方法 | 是否端到端训练 | A-D | A-W | D-A | ||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
OS* | UNK | HOS | OS* | UNK | HOS | OS* | UNK | HOS | ||||
DANN[ | √ | 94.0 | — | — | 93.5 | — | — | 84.6 | — | — | ||
AMS[ | √ | 89.3 | 71.8 | 79.6 | 87.3 | 67.3 | 76.0 | 74.9 | 65.4 | 69.8 | ||
DAMC[ | √ | 19.1 | 84.0 | 31.1 | 25.2 | 88.9 | 39.3 | 21.9 | 92.3 | 35.4 | ||
DCC[ | √ | 91.3 | 41.3 | 56.9 | 92.3 | 43.6 | 59.2 | 76.2 | 70.0 | 73.0 | ||
文献[ | × | 89.1 | 52.6 | 66.2 | 85.7 | 43.8 | 57.9 | 76.3 | 52.8 | 62.4 | ||
文献[ | √ | 88.4 | 76.2 | 81.8 | 81.6 | 82.0 | 81.8 | 58.2 | 92.6 | 71.5 | ||
方法 | D-W | W-A | W-D | Avg | ||||||||
OS* | UNK | HOS | OS* | UNK | HOS | OS* | UNK | HOS | OS* | UNK | HOS | |
DANN[ | 100.0 | — | — | 78.1 | — | — | 100.0 | — | — | 91.7 | — | — |
AMS[ | 99.3 | 78.9 | 88.0 | 76.8 | 70.2 | 73.3 | 100.0 | 82.8 | 90.6 | 88.0 | 72.8 | 79.6 |
DAMC[ | 65.0 | 94.1 | 76.9 | 25.1 | 88.6 | 39.1 | 75.3 | 91.9 | 82.8 | 38.6 | 90.0 | 50.8 |
DCC[ | 100.0 | 66.0 | 79.5 | 72.5 | 60.8 | 66.2 | 100.0 | 66.0 | 79.5 | 88.7 | 59.5 | 70.1 |
文献[ | 100.0 | 34.8 | 51.7 | 77.0 | 49.1 | 60.0 | 100.0 | 46.2 | 63.2 | 88.0 | 46.6 | 60.2 |
文献[ | 99.2 | 93.4 | 96.2 | 60.6 | 92.5 | 73.2 | 100.0 | 93.0 | 96.4 | 81.3 | 88.3 | 83.5 |
方法 | 是否端到端训练 | Rw-Pr | Rw-Cl | Pr-Rw | ||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
OS* | UNK | HOS | OS* | UNK | HOS | OS* | UNK | HOS | ||||
OSBP[ | √ | 66.45 | 54.89 | 60.12 | 41.89 | 58.49 | 48.82 | 66.19 | 60.80 | 63.38 | ||
ROS[ | × | 66.60 | 39.52 | 49.61 | 43.53 | 44.10 | 43.81 | 63.89 | 47.23 | 54.43 | ||
DANCE[ | √ | 66.32 | 57.18 | 61.41 | 42.40 | 67.63 | 52.12 | 64.07 | 64.43 | 64.25 | ||
OMEGA[ | √ | 59.61 | 77.89 | 67.53 | 42.23 | 69.58 | 52.56 | 62.75 | 73.80 | 67.83 | ||
OVA[ | √ | 67.34 | 60.23 | 63.58 | 35.66 | 77.59 | 48.87 | 63.51 | 74.00 | 68.36 | ||
方法 | Pr-Cl | Cl-Rw | Cl-Pr | Avg | ||||||||
OS* | UNK | HOS | OS* | UNK | HOS | OS* | UNK | HOS | OS* | UNK | HOS | |
OSBP[ | 38.03 | 59.67 | 46.45 | 59.47 | 53.73 | 56.45 | 55.56 | 60.48 | 57.91 | 54.60 | 58.01 | 55.52 |
ROS[ | 41.53 | 45.05 | 43.18 | 58.57 | 34.42 | 43.49 | 51.85 | 26.42 | 35.13 | 54.33 | 39.46 | 44.94 |
DANCE[ | 46.70 | 66.27 | 54.79 | 57.09 | 62.14 | 59.51 | 51.07 | 68.50 | 58.51 | 54.61 | 64.36 | 58.43 |
OMEGA[ | 45.81 | 69.81 | 55.32 | 57.35 | 68.83 | 62.57 | 58.49 | 64.04 | 61.14 | 54.37 | 70.66 | 61.16 |
OVA[ | 32.70 | 83.73 | 47.03 | 51.25 | 78.39 | 61.98 | 46.86 | 72.30 | 56.86 | 49.55 | 74.37 | 57.78 |
表4 在Office-Home数据集上各OSDA方法的结果 (%)
Tab. 4 Results of various OSDA methods on Office-Home dataset
方法 | 是否端到端训练 | Rw-Pr | Rw-Cl | Pr-Rw | ||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
OS* | UNK | HOS | OS* | UNK | HOS | OS* | UNK | HOS | ||||
OSBP[ | √ | 66.45 | 54.89 | 60.12 | 41.89 | 58.49 | 48.82 | 66.19 | 60.80 | 63.38 | ||
ROS[ | × | 66.60 | 39.52 | 49.61 | 43.53 | 44.10 | 43.81 | 63.89 | 47.23 | 54.43 | ||
DANCE[ | √ | 66.32 | 57.18 | 61.41 | 42.40 | 67.63 | 52.12 | 64.07 | 64.43 | 64.25 | ||
OMEGA[ | √ | 59.61 | 77.89 | 67.53 | 42.23 | 69.58 | 52.56 | 62.75 | 73.80 | 67.83 | ||
OVA[ | √ | 67.34 | 60.23 | 63.58 | 35.66 | 77.59 | 48.87 | 63.51 | 74.00 | 68.36 | ||
方法 | Pr-Cl | Cl-Rw | Cl-Pr | Avg | ||||||||
OS* | UNK | HOS | OS* | UNK | HOS | OS* | UNK | HOS | OS* | UNK | HOS | |
OSBP[ | 38.03 | 59.67 | 46.45 | 59.47 | 53.73 | 56.45 | 55.56 | 60.48 | 57.91 | 54.60 | 58.01 | 55.52 |
ROS[ | 41.53 | 45.05 | 43.18 | 58.57 | 34.42 | 43.49 | 51.85 | 26.42 | 35.13 | 54.33 | 39.46 | 44.94 |
DANCE[ | 46.70 | 66.27 | 54.79 | 57.09 | 62.14 | 59.51 | 51.07 | 68.50 | 58.51 | 54.61 | 64.36 | 58.43 |
OMEGA[ | 45.81 | 69.81 | 55.32 | 57.35 | 68.83 | 62.57 | 58.49 | 64.04 | 61.14 | 54.37 | 70.66 | 61.16 |
OVA[ | 32.70 | 83.73 | 47.03 | 51.25 | 78.39 | 61.98 | 46.86 | 72.30 | 56.86 | 49.55 | 74.37 | 57.78 |
[1] | PAN S J, YANG Q. A survey on transfer learning [J]. IEEE Transactions on Knowledge and Data Engineering, 2010, 22(10): 1345-1359. |
[2] | DAUMÉ H, Ⅲ. Frustratingly easy domain adaptation [C]// Proceedings of the 45th Annual Meeting of the Association of Computational Linguistics. Stroudsburg: ACL, 2007: 256-263. |
[3] | ZHANG L, GAO X. Transfer adaptation learning: a decade survey[J]. IEEE Transactions on Neural Networks and Learning Systems, 2024, 35(1): 23-44. |
[4] | GANIN Y, USTINOVA E, AJAKAN H, et, al. Domain-adversarial training of neural networks [J]. Journal of Machine Learning Research, 2016, 17: 1-35. |
[5] | WANG M, DENG W. Deep visual domain adaptation: a survey[J]. Neurocomputing, 2018, 312: 135-153. |
[6] | LUO Y, ZHENG L, GUAN T, et al. Taking a closer look at domain shift: category-level adversaries for semantics consistent domain adaptation [C]// Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2019: 2502-2511. |
[7] | BUSTO P P, GALL J. Open set domain adaptation [C]// Proceedings of the 2017 IEEE International Conference on Computer Vision. Piscataway: IEEE, 2017: 754-763. |
[8] | SCHEIRER W J, DE REZENDE ROCHA A, SAPKOTA A, et al. Toward open set recognition [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2013, 35(7): 1757-1772. |
[9] | SAITO K, YAMAMOTO S, USHIKU Y, et al. Open set domain adaptation by backpropagation [C]// Proceedings of the 2018 European Conference on Computer Vision, LNCS 11209. Cham: Springer, 2018:156-171. |
[10] | GENG C, HUANG S J, CHEN S. Recent advances in open set recognition: a survey [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021, 43(10): 3614-3631. |
[11] | PAWSON R, WONG G, OWEN L. Known knowns, known unknowns, unknown unknowns: the predicament of evidence-based policy [J]. American Journal of Evaluation, 2011, 32(4): 518-546. |
[12] | SUGIYAMA M, KRAULEDAT M, MÜLLER K R. Covariate shift adaptation by importance weighted cross validation [J]. Journal of Machine Learning Research, 2007, 8: 985-1005. |
[13] | GHIFARY M, KLEIJN W B, ZHANG M. Domain adaptive neural networks for object recognition [C]// Proceedings of the 2014 Pacific Rim International Conference on Artificial Intelligence, LNCS 8862. Cham: Springer, 2014:898-904. |
[14] | GOPALAN R, LI R, CHELLAPPA R. Domain adaptation for object recognition: an unsupervised approach [C]// Proceedings of the 2011 IEEE International Conference on Computer Vision. Piscataway: IEEE, 2011: 999-1006. |
[15] | LONG M, CAO Z, WANG J, et al. Conditional adversarial domain adaptation [C]// Proceedings of the 32nd International Conference on Neural Information Processing Systems. Red Hook: Curran Associates Inc., 2018: 1647-1657. |
[16] | 范苍宁,刘鹏,肖婷,等. 深度域适应综述:一般情况与复杂情况[J]. 自动化学报, 2021, 47(3):515-548. |
FAN C N, LIU P, XIAO T, et al. A review of deep domain adaptation: general situation and complex situation [J]. Acta Automatica Sinica, 2021, 47(3): 515-548. | |
[17] | FANG Z, LU J, LIU F, et al. Open set domain adaptation: theoretical bound and algorithm [J]. IEEE Transactions on Neural Networks and Learning Systems, 2021, 32(10): 4309-4322. |
[18] | ZHONG L, FANG Z, LIU F, et al. Bridging the theoretical bound and deep algorithms for open set domain adaptation [J]. IEEE Transactions on Neural Networks and Learning Systems, 2023, 34(8): 3859-3873. |
[19] | TAN C, SUN F, KONG T, et al. A survey on deep transfer learning [C]// Proceedings of the 2018 International Conference on Artificial Neural Networks, LNCS 11141. Cham: Springer, 2018: 270-279. |
[20] | BUCCI S, LOGHMANI M R, TOMMASI T. On the effectiveness of image rotation for open set domain adaptation [C]// Proceedings of the 2020 European Conference on Computer Vision, LNCS 12361. Cham: Springer, 2020: 422-438. |
[21] | GIDARIS S, SINGH P, KOMODAKIS N. Unsupervised representation learning by predicting image rotations [EB/OL]. [2024-07-11]. . |
[22] | KUNDU JN, VENKAT N, REVANUR A, et al. Towards inheritable models for open-set domain adaptation [C]// Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2020: 12373-12382. |
[23] | ZEILER M D, FERGUS R. Visualizing and Understanding convolutional networks [C]// Proceedings of the 2014 European Conference on Computer Vision, LNCS 8689. Cham: Springer, 2014: 818-833. |
[24] | CHEN H, WANG Y, XU C, et al. Data-free learning of student networks [C]// Proceedings of the 2019 IEEE/CVF International Conference on Computer Vision. Piscataway: IEEE, 2019: 3513-3521. |
[25] | KUNDU JN, VENKAT N, RAHUL M V, et al. Universal source-free domain adaptation [C]// Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2020: 4543-4552. |
[26] | BAKTASHMOTLAGH M, CHEN T, SALZMANN M. Learning to generate the unknowns as a remedy to the open-set domain shift[C]// Proceedings of the 2022 IEEE/CVF Winter Conference on Applications of Computer Vision. Piscataway: IEEE, 2022: 3737-3746. |
[27] | LIU Y, DENG A, DENG M, et al. Transforming the open set into a pseudo-closed set: a regularized GAN for domain adaptation in open-set fault diagnosis [J]. IEEE Transactions on Instrumentation and Measurement, 2023, 72: No.3531312. |
[28] | XU Q, SHI Y, YUAN X, et al. Universal domain adaptation for remote sensing image scene classification [J]. IEEE Transactions on Geoscience and Remote Sensing, 2023, 61: No.4700515. |
[29] | LIU J, JING M, LI J, et al. Open set domain adaptation via joint alignment and category separation [J]. IEEE Transactions on Neural Networks and Learning Systems, 2023, 34(9): 6186-6199. |
[30] | MENDES JÚNIOR P R, DE SOUZA R M, WERNECK R D O, et al. Nearest neighbors distance ratio open-set classifier [J]. Machine Learning, 2017, 106(3): 359-386. |
[31] | JING M, LI J, ZHU L, et al. Balanced open set domain adaptation via centroid alignment [C]// Proceedings of the 35th AAAI Conference on Artificial Intelligence. Palo Alto: AAAI Press, 2021: 8013-8020. |
[32] | DAVIDSON T R, FALORSI L, DE CAO N, et al. Hyperspherical variational auto-encoders [C]// Proceedings of the 34th Conference on Uncertainty in Artificial Intelligence. Arlington, VA: AUAI Press, 2018: 856-865. |
[33] | LI X, LI J, DU Z, et al. Interpretable open-set domain adaptation via angular margin separation [C]// Proceedings of the 2022 European Conference on Computer Vision, LNCS 13694. Cham: Springer, 2022: 1-18. |
[34] | KRICHEN M. Generative adversarial networks [C]// Proceedings of the 14th International Conference on Computing Communication and Networking Technologies. Piscataway: IEEE, 2023: 1-7. |
[35] | GANIN Y, LEMPITSKY V. Unsupervised domain adaptation by backpropagation [C]// Proceedings of the 32nd International Conference on Machine Learning. New York: JMLR.org, 2015: 1180-1189. |
[36] | FU J, WU X, ZHANG S, et al. Improved open set domain adaptation with backpropagation [C]// Proceedings of the 2019 IEEE International Conference on Image Processing. Piscataway: IEEE, 2019: 2506-2510. |
[37] | SHERMIN T, LU G, TENG S W, et al. Adversarial network with multiple classifiers for open set domain adaptation [J]. IEEE Transactions on Multimedia, 2021, 23: 2732-2744. |
[38] | GAO Y, MA A J, GAO Y, et al. Adversarial open set domain adaptation via progressive selection of transferable target samples[J]. Neurocomputing, 2020, 410: 174-184. |
[39] | ZHANG H J, LI A, GUO J, et al. Improving open set domain adaptation using image-to-image translation and instance-weighted adversarial learning [J]. Journal of Computer Science and Technology, 2023, 38(3): 644-658. |
[40] | FENG Q, KANG G, FAN H, et al. Attract or distract: exploit the margin of open set [C]// Proceedings of the 2019 IEEE/CVF International Conference on Computer Vision. Piscataway: IEEE, 2019: 7989-7998. |
[41] | QI C, SU F. Contrastive-center loss for deep neural networks [C]// Proceedings of the 2017 IEEE International Conference on Image Processing. Piscataway: IEEE, 2017: 2851-2855. |
[42] | JING T, LIU H, DING Z. Towards novel target discovery through open-set domain adaptation [C]// Proceedings of the 2021 IEEE/CVF International Conference on Computer Vision. Piscataway: IEEE, 2021: 9302-9311. |
[43] | PAN Y, YAO T, LI Y, et al. Exploring category-agnostic clusters for open-set domain adaptation [C]// Proceedings of the 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2020: 13864-13872. |
[44] | HJELM R D, FEDOROV A, LAVOIE-MARCHILDON S, et al. Learning deep representations by mutual information estimation and maximization [EB/OL]. [2024-07-14].. |
[45] | LI G, KANG G, ZHU Y, et al. Domain consensus clustering for universal domain adaptation [C]// Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2021: 9752-9761. |
[46] | SAITO K, KIM D, SCLAROFF S, et al. Universal domain adaptation through self supervision [C]// Proceedings of the 34th International Conference on Neural Information Processing Systems. Red Hook: Curran Associates Inc., 2020: 16282-16292. |
[47] | RU J, TIAN J, XIAO C, et al. Imbalanced open set domain adaptation via moving-threshold estimation and gradual alignment[J]. IEEE Transactions on Multimedia, 2024, 26: 2504-2514. |
[48] | SAITO K, WATANABE K, USHIKU Y, et al. Maximum classifier discrepancy for unsupervised domain adaptation [C]// Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2018: 3723-3732. |
[49] | XU Y, CHEN L, DUAN L, et al. Open set domain adaptation with soft unknown-class rejection [J]. IEEE Transactions on Neural Networks and Learning Systems, 2023, 34(3): 1601-1612. |
[50] | GAO F, PI D, CHEN J. Balanced and robust unsupervised open set domain adaptation via joint adversarial alignment and unknown class isolation [J]. Expert Systems with Applications, 2024, 238(Pt E): No.122127. |
[51] | SAITO K, SAENKO K. OVANet: one-vs-all network for universal domain adaptation [C]// Proceedings of the 2021 IEEE/CVF International Conference on Computer Vision. Piscataway: IEEE, 2021: 8980-8989. |
[52] | SILVA L F A E, SEBE N, ALMEIDA J. Tightening classification boundaries in open set domain adaptation through unknown exploitation [C]// Proceedings of the 36th SIBGRAPI Conference on Graphics, Patterns and Images. Piscataway: IEEE, 2023: 157-162. |
[53] | PENG X, USMAN B, KAUSHIK N, et al. VisDA: the visual domain adaptation challenge [EB/OL]. [2025-01-14]. . |
[54] | SAENKO K, KULIS B, FRITZ M, et al. Adapting visual category models to new domains [C]// Proceedings of the 2010 European Conference on Computer Vision, LNCS 6314. Berlin: Springer, 2010: 213-226. |
[55] | VENKATESWARA H, EUSEBIO J, CHAKRABORTY S, et al. Deep hashing network for unsupervised domain adaptation[C]// Proceedings of the 2017 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2017: 5385-5394. |
[56] | YOU K, LONG M, CAO Z, et al. Universal domain adaptation[C]// Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE, 2019: 2715-2724. |
[1] | 何玉林, 何芃, 黄哲学, 解为成, PHILIPPE Fournier-Viger. 以标注确定性增强为导向的正类-无标签学习算法[J]. 《计算机应用》唯一官方网站, 2025, 45(7): 2101-2112. |
[2] | 冯博, 于海征, 边红. 基于掩码增强自训练的域适应语义分割[J]. 《计算机应用》唯一官方网站, 2025, 45(7): 2132-2137. |
[3] | 孙雨阳, 张敏婕, 胡婕. 基于语义前缀微调的零样本对话状态跟踪领域迁移模型[J]. 《计算机应用》唯一官方网站, 2025, 45(7): 2221-2228. |
[4] | 姜超英, 李倩, 刘宁, 刘磊, 崔立真. 基于图对比学习的再入院预测模型[J]. 《计算机应用》唯一官方网站, 2025, 45(6): 1784-1792. |
[5] | 崔双双, 王宏志, 朱加昊, 吴昊. 面向低能耗高性能的分类器两阶段数据选择方法[J]. 《计算机应用》唯一官方网站, 2025, 45(6): 1703-1711. |
[6] | 李道全, 徐正, 陈思慧, 刘嘉宇. 融合变分自编码器与自适应增强卷积神经网络的网络流量分类模型[J]. 《计算机应用》唯一官方网站, 2025, 45(6): 1841-1848. |
[7] | 李雪莹, 杨琨, 涂国庆, 刘树波. 基于局部增强的时序数据对抗样本生成方法[J]. 《计算机应用》唯一官方网站, 2025, 45(5): 1573-1581. |
[8] | 田仁杰, 景明利, 焦龙, 王飞. 基于混合负采样的图对比学习推荐算法[J]. 《计算机应用》唯一官方网站, 2025, 45(4): 1053-1060. |
[9] | 孙海涛, 林佳瑜, 梁祖红, 郭洁. 结合标签混淆的中文文本分类数据增强技术[J]. 《计算机应用》唯一官方网站, 2025, 45(4): 1113-1119. |
[10] | 丁美荣, 卓金鑫, 陆玉武, 刘庆龙, 郎济聪. 融合环境标签平滑与核范数差异的领域自适应[J]. 《计算机应用》唯一官方网站, 2025, 45(4): 1130-1138. |
[11] | 陈庆礼, 郭渊博, 方晨. 面向数据异构的聚类联邦学习算法[J]. 《计算机应用》唯一官方网站, 2025, 45(4): 1086-1094. |
[12] | 盛坤, 王中卿. 基于大语言模型和数据增强的通感隐喻分析[J]. 《计算机应用》唯一官方网站, 2025, 45(3): 794-800. |
[13] | 孙晨伟, 侯俊利, 刘祥根, 吕建成. 面向工程图纸理解的大语言模型提示生成方法[J]. 《计算机应用》唯一官方网站, 2025, 45(3): 801-807. |
[14] | 王瑜, 方贤进, 杨高明, 丁一峰, 杨新露. 基于注意力掩码与特征提取的人脸伪造主动防御[J]. 《计算机应用》唯一官方网站, 2025, 45(3): 904-910. |
[15] | 富坤, 应世聪, 郑婷婷, 屈佳捷, 崔静远, 李建伟. 面向小样本节点分类的图数据增强方法[J]. 《计算机应用》唯一官方网站, 2025, 45(2): 392-402. |
阅读次数 | ||||||
全文 |
|
|||||
摘要 |
|
|||||