|
|
FCS 前沿研究:用于无监督域适应的自适应标签过滤学习 |
|
论文标题:Self-adaptive label filtering learning for unsupervised domain adaptation(用于无监督域适应的自适应标签过滤学习)
期刊:Frontiers of Computer Science
作者:Qing TIAN, Heyang SUN, Shun PENG, Tinghuai MA
发表时间:15 Feb 2023
DOI:10.1007/s11704-022-1283-6
微信链接:点击此处阅读微信文章
原文信息
标 题:
Self-adaptive label filtering learning for unsupervised domain adaptation
发表年份:
2023年
原文链接:
https://journal.hep.com.cn/fcs/EN/10.1007/s11704-022-1283-6
引用格式:
Qing TIAN, Heyang SUN, Shun PENG, Tinghuai MA. Self-adaptive label filtering learning for unsupervised domain adaptation. Front. Comput. Sci., 2023, 17(1): 171308
公众号推文链接:
文章精要 | 南京信息工程大学田青教授团队:用于无监督域适应的自适应标签过滤学习
01 导读
作为一种新的机器学习范式,无监督域适应(Unsupervised Domain Adaptation, UDA)旨在利用相关但分布不同的源域(Source Domain)的知识来辅助训练一个对无标记的目标域(Target Domain)有效的模型。现有大多数无监督域适应方法通过预训练模型来获取目标域的伪标签从而进行类级别的分布对齐,但现有的为标签编码的方式中,硬标签过度置信而软标签往往含有噪声,都容易影响模型的表现。因此,如图1所示,本篇论文提出了一种自适应的标签过滤学习(SALFL)方法来解决这个问题。具体来说,本篇论文设计了一种基于图的随机游走策略来预测伪标签然后通过所提出的自适应标签过滤机制对其进行提炼。更进一步地,在新获得的伪标签基础之上,本篇论文提出了一种更通用的联合分布对齐方法并将整体方法推广到深度算法框架中。
![](data:image/png;base64,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)
图1. SALFL框架示意图。概率标签上不同颜色表示对应不同的类别。
02 相关工作
如导读中所述,域适应问题的关键之处在于寻找源域和目标域之间的相似性,从而利用源域的知识来辅助目标域的学习,因此两类方法得到了较多的关注。
基于特征的统计对齐方法是一种较为有效的方法,举例来说,测地线流核(GFK)方法[1]通过度量最优维数将所得跨域子空间进行积分,以从统计方法上建模域偏移。尽管该方法优雅且有效,但灵活性不足难以进行拓展。之后,子空间对齐(SA)[2]方法将源域特征转换至目标域空间,这种方式在一定程度上对齐了域偏移但忽略了目标域空间的私有部分。实际上,大多数现有UDA方法试图通过共享子空间来消除域差异。基于此,最大均值差异嵌入(MMDE)[3]方法对MMD进行最大方差展开(MVU),从而得到一个对齐后的低维潜在子空间。类似地,迁移成分分析(TCA)[4]方法通过MMD最小化源域和目标域的散点,将其投影至公共空间中。除了总体分布对齐外,联合域适应(JDA)[5]方法将边缘MMD与条件MMD进行结合以更好地进行域对齐。为了解决UDA中的任务不平衡问题,迁移联合匹配(TJM)[6]方法对跨域样本进行重新加权并进行联合特征匹配。后来,平衡域适应(BDA)[7]方法通过在边际分布和条件分布对齐之间引入一个平衡因子,获得了比TJM更好的结果。为了利用更多的判别知识,联合几何和统计对齐(JGSA)[8]方法将目标域、源域的分布方差以及跨域分布散差联合建模为一个统一的目标。域不变类判别(DICD)[9]方法通过测量类内相似性和类间差异性来学习UDA的类判别表示。
除此之外,几何结构学习因在数据结构挖掘上的有效性,在UDA任务中被广泛使用。例如,流行嵌入分布对齐(MEDA)[10]方法被设计为通过最小化格拉斯曼流形上的数据结构风险来学习一个领域不变的UDA分类器。判别性结构意识域适应(DGA-DA)[11]方法通过几何结构学习推断目标域标签来执行UDA。然而由于几何结构学习往往通过迭代优化策略实现,所以每轮生成的伪标签质量对下一轮迭代至关重要,低置信度的伪标签很容易造成累计误差从而影响整个模型的性能表现。不幸的是,这一问题在上述讨论的方法中均被忽略。
实际上,在无监督域适应中,伪标签的问题也得到了一定的关注。具体来说,置信正则自训练(CRST)[12]方法通过置信正则化来提升软伪标签的平滑性,从而获得更可信的预测,但所嵌入的正则化损失难以优化,也难以转移到其他模型。此外,对比适应网络(CAN)[13]方法通过置零来过滤远离集群中心的点和类。虽然该方法可以减轻噪声伪标签的干扰,但它存在伪标签过置信问题,而这篇论文提出的SALFL方法可以同时处理这两个问题。渐进特征对齐网络(PFAN)[14]方法利用从易到难的转移策略(EHTS)逐步选择简单的样本进行域对齐。虽然每个样本的置信度保持不变,但由于所选样本组成的局部域具有较高的置信度,更易于对齐。与这类迭代和局部选择样本的策略不同,这篇论文根据所提自适应标签滤波学习策略对每个目标域样本都进行有效的提纯从而使所有样本都更加可靠。
03 主要方法
问题定义
在UDA场景中,给定一个有标签的源域数据集
,其中,
表示第
个源域样本实例,
表示其对应的标签。UDA的目标是为无标记的目标域
样本
分配标签,其中,
表示第
个目标域样本实例。源域和目标域共享相同的特征空间
和标记空间
。除此之外,每个域的边缘和条件分布均不相同,即
和
。
基于图的随机游走
这篇论文首先引入了一种标签预测算法,它在图上采用一种特殊的随机游走的方式,该算法允许未标记点(即目标域数据)在邻域相似度图的指导下进行随机游走,从标记点(即源域数据)中寻找标签信息。该图结构的邻接矩阵
的每个元素可计算如下:
![](data:image/png;base64,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)
其中
度量
和
之间的非负相似性,
表示方差,表示域中最相近的
个近邻组成的集合。每一次的随机游走变换可以计算如下:
![](data:image/png;base64,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)
其中
表示近邻图的路径权重矩阵计算,度矩阵D是一个每个元素为
的对角阵。除此之外,对角矩阵
和
各自由方向系数
和
构成。如果第
个点无标记(即目标域样本),定义
这样可以使得这个点进行随机游走,反之如果这个点有标记(即源域样本),定义
这样可以约束这个点停在原地。
在近邻图中,每个点根据变换矩阵
从起始点进行随机行走,直到其连续两次到达同一点。定义
,
表示在第
个点在第
个点上停止随机游走。那么,随机行走的过程可以描述为:
![](data:image/png;base64,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)
其中
表示第
个点在第
步停留在第
点上的概率。更进一步地,软标签矩阵
可以被计算为
。
自适应标签过滤学习
通过上述的随机游走过程得到的目标域预测标签对后续的域分布对齐至关重要。下一步是选择用硬标签或软标签来进行标签编码。然而,硬标签过度自信,当出现误分类样本时很容易误导模型,导致负迁移。同样地,软标签所携带的噪声标签也容易造成模型的混淆。为了克服这一问题,本文通过自适应标签过滤学习来滤除目标域的错误或有噪声伪标签。令
分别表示源域和目标域的特征,
表示通过随机行走预测得到的第i个目标域实例的伪标签编码向量,同时
为其属于第
个类的概率。首先,通过最近邻分类器为目标域的每个实例
重新分配标记:
![](data:image/png;base64,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)
其中
表示两点之间的欧氏距离,
表示源域第
个类的类中心。然后,按照如下方式更新(提纯)伪标签:
![](data:image/png;base64,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)
其中,当
时,单位阶跃函数
,否则为0。定义函数非零
,令它返回向量
中的非零值的数量。如图2所示,对于一个由自适应标签过滤策略所提纯后的标签
,如果
,则认为它为置信标签(如图2(a)),该标签可以为后续类级别对齐提供更准确的分类信息。如果
,则认为它为模糊标签(如图2(b)),虽然无法提供更准确的分类信息,但该标签可以有效避免误分类造成的累计误差,并且免除了其它微小噪声的干扰。本章将所提的提纯过程叫做自适应标签过滤,它利用源域分布的质心来近似目标域的质心,是一种目标域无监督方式。实际上,这种自适应标签过滤学习过程可以看作是基于标签选择的软标签编码和硬标签编码的自动组合,置信的标签使用硬标签的形式,而模糊的标签使用软标签的形式,改策略可以有效地缓解负迁移的产生。
![](data:image/png;base64,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)
图2. 过滤后的伪标签示意图。
基于过滤标签的域适应一般形式
为了进行域分布对齐,本篇论文基于MMD进行联合分布对齐,其边缘分布对齐可以定义为
,其中
是子空间投影矩阵。令
,
,
,则MMD矩阵为
。与其他方法不同的是,为了有效地将过滤后的标签与条件MMD度量相结合,本篇论文重新定义了跨域的条件分布距离度量为
,其中
,
,
和
分别是源域和目标域通过随机游走获得的标签。重新定义的条件MMD矩阵
的每个元素可以计算如下:
![](data:image/png;base64,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)
虽然联合域分布对齐可以增加模型对目标域的判别性,但它只关注来自同一类别的数据,而忽略了来自相同域的不同类间的可判别性。为此,本篇论文最大化这两个领域的不同类别集合之间的可区分性。具体来说,在联合域
中,用
表示第
个类的组成的子域,
表示除第
个类外的所有其它类别的组成的子域。然后,类别间的差异判别可以表达为
,其中,
表示判别矩阵,它的每个元素定义为:
![](data:image/png;base64,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)
最大化
可以有效地扩大源域和目标域的类别之间的距离,进一步提高了模型的判别能力。
总体目标函数及优化
综合考虑上述所有组成部分,SALFL方法的最终目标可以整合如下:
![](data:image/png;base64,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)
其中
为超参数,
为中心矩阵。整体目标函数的优化由两部分组成,首先是通过基于图的随机游走进行标签预测:
![](data:image/png;base64,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)
第二个需要优化的部分是基于标签过滤的判别性域适应。将主成分分析(PCA)约束嵌入增广拉格朗日方法中,最终目标函数可以被改写为
![](data:image/png;base64,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)
最终,该公式可以通过广义特征值分解求解,最优投影矩阵由其前个最小的特征向量构成。
04 实验结果
为了验证SALFL模型的有效性,作者在Office-Caltech-10、Office-Home、Office-31等域适应标准数据集上进行了实验,其实验结果如下:
表1. 在Office-Caltech-10数据集上使用SURF特征的分类性能。
![](/upload/paper/images/2023/5/2023561434331020.png)
表2. 在Office-Caltech-10数据集上使用图片特征的分类性能。
![](/upload/paper/images/2023/5/202356143554610.png)
表3. 在Office-Home数据集上使用ResNet-50特征的分类性能。
![](/upload/paper/images/2023/5/2023561435287890.png)
可以看出,SALFL方法在不同的数据集、不同的特征类型上,相较于其他方法其准确率有实质性的提升。除此之外,这篇论文所提的自适应的标签过滤学习方法可以与大多数现有域适应方法结合使用,其与深度方法结合使用的结果如下所示:
表4. 在Office-31数据集上与不同深度方法相结合的分类性能。
![](/upload/paper/images/2023/5/2023561435465080.png)
可以发现,现有域适应方法在与所提标签过滤学习方法相结合后,性能较原模型有了一定的提升。为了进一步验证模型的有效性,作者还进行了假设检验,如图3所示,可以发现所提SALFL方法相较其他方法有一定的显著性。
![](/upload/paper/images/2023/5/202356143625240.png)
图3. 在Office-31数据集上与不同深度方法相结合的分类性能。
参考文献
[1] Gong B, Shi Y, Sha F, et al. Geodesic flow kernel for unsupervised domain adaptation. 2012 IEEE conference on computer vision and pattern recognition. IEEE, 2012: 2066-2073.
[2] Fernando B, Habrard A, Sebban M, et al. Unsupervised visual domain adaptation using subspace alignment. Proceedings of the IEEE international conference on computer vision. 2013: 2960-2967.
[3] Pan S J, Kwok J T, Yang Q. Transfer learning via dimensionality reduction. AAAI. 2008, 8: 677-682.
[4] Pan S J, Tsang I W, Kwok J T, et al. Domain adaptation via transfer component analysis. IEEE transactions on neural networks, 2010, 22(2): 199-210.
[5] Long M, Wang J, Ding G, et al. Transfer feature learning with joint distribution adaptation. Proceedings of the IEEE international conference on computer vision. 2013: 2200-2207.
[6] Long M, Wang J, Ding G, et al. Transfer joint matching for unsupervised domain adaptation. Proceedings of the IEEE conference on computer vision and pattern recognition. 2014: 1410-1417.
[7] Wang J, Chen Y, Hao S, et al. Balanced distribution adaptation for transfer learning. 2017 IEEE international conference on data mining (ICDM). IEEE, 2017: 1129-1134.
[8] Zhang J, Li W, Ogunbona P. Joint geometrical and statistical alignment for visual domain adaptation. Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 1859-1867.
[9] Li S, Song S, Huang G, et al. Domain invariant and class discriminative feature learning for visual domain adaptation. IEEE transactions on image processing, 2018, 27(9): 4260-4273.
[10] Wang J, Feng W, Chen Y, et al. Visual domain adaptation with manifold embedded distribution alignment. Proceedings of the 26th ACM international conference on Multimedia. 2018: 402-410.
[11] Luo L, Chen L, Hu S, et al. Discriminative and geometry-aware unsupervised domain adaptation. IEEE transactions on cybernetics, 2020, 50(9): 3914-3927.
[12] Zou Y, Yu Z, Liu X, et al. Confidence regularized self-training. Proceedings of the IEEE/CVF International Conference on Computer Vision. 2019: 5982-5991.
[13] Kang G, Jiang L, Yang Y, et al. Contrastive adaptation network for unsupervised domain adaptation. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2019: 4893-4902.
[14] Chen C, Xie W, Huang W, et al. Progressive feature alignment for unsupervised domain adaptation. Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2019: 627-636.
解读:徐 宁 东南大学
审核:张 琨 合肥工业大学
![](/upload/paper/images/2023/5/202356143643860.png)
Frontiers of Computer Science
Frontiers of Computer Science (FCS)是由教育部主管、高等教育出版社和北京航空航天大学共同主办、SpringerNature 公司海外发行的英文学术期刊。本刊于 2007 年创刊,双月刊,全球发行。主要刊登计算机科学领域具有创新性的综述论文、研究论文等。本刊主编为周志华教授,共同主编为熊璋教授。编委会及青年 AE 团队由国内外知名学者及优秀青年学者组成。本刊被 SCI、Ei、DBLP、INSPEC、SCOPUS 和中国科学引文数据库(CSCD)核心库等收录,为 CCF 推荐期刊;两次入选“中国科技期刊国际影响力提升计划”;入选“第4届中国国际化精品科技期刊”;入选“中国科技期刊卓越行动计划项目”。
《前沿》系列英文学术期刊
由教育部主管、高等教育出版社主办的《前沿》(Frontiers)系列英文学术期刊,于2006年正式创刊,以网络版和印刷版向全球发行。系列期刊包括基础科学、生命科学、工程技术和人文社会科学四个主题,是我国覆盖学科最广泛的英文学术期刊群,其中12种被SCI收录,其他也被A&HCI、Ei、MEDLINE或相应学科国际权威检索系统收录,具有一定的国际学术影响力。系列期刊采用在线优先出版方式,保证文章以最快速度发表。
中国学术前沿期刊网
http://journal.hep.com.cn
![](/upload/paper/images/2023/5/2023561437148830.png)
特别声明:本文转载仅仅是出于传播信息的需要,并不意味着代表本网站观点或证实其内容的真实性;如其他媒体、网站或个人从本网站转载使用,须保留本网站注明的“来源”,并自负版权等法律责任;作者如果不希望被转载或者联系转载稿费等事宜,请与我们接洽。