![迁移学习算法:应用与实践](https://wfqqreader-1252317822.image.myqcloud.com/cover/428/47755428/b_47755428.jpg)
上QQ阅读APP看本书,新人免费读10天
设备和账号都新为新人
![](https://epubservercos.yuewen.com/A97C42/27167652804832906/epubprivate/OEBPS/Images/3_01.jpg?sign=1734436936-XHopR4Q8mO0pKcu7PSq1AGeC9p4fWXPI-0-f9bb24dc9a3b365ae13cbdeacd363ac3)
图4.5 表达图像完整与部分信息的示例
![](https://epubservercos.yuewen.com/A97C42/27167652804832906/epubprivate/OEBPS/Images/3_02.jpg?sign=1734436936-8e65yOpJPOCfA7HHsOXWNbHVfYcagYku-0-0db5307aa1c06743ac4ff4e766cd80dd)
图4.7 单源领域自适应与多源领域自适应。在单源领域适应中,源领域和目标领域的分布不能很好地匹配,而在多源领域适应中,由于多个源领域之间的分布偏移,匹配所有源领域和目标领域的分布要困难得多[71]
![](https://epubservercos.yuewen.com/A97C42/27167652804832906/epubprivate/OEBPS/Images/3_03.jpg?sign=1734436936-rjWsgmGvWzBDl4psl7WNpAWUoA6RvIQb-0-dd8b2b2f455552c4c4ec24a8e9c99e12)
图4.8 同时对齐分布和分类器的多源自适应方法[71]
![](https://epubservercos.yuewen.com/A97C42/27167652804832906/epubprivate/OEBPS/Images/4_01.jpg?sign=1734436936-1jhMUk4lmGWAK8YHsHQF8Xe469238Ymh-0-4b10a026d6e0525a76d2d5630d6ad541)
图5.4 领域对抗神经网络可视化结果[64]
![](https://epubservercos.yuewen.com/A97C42/27167652804832906/epubprivate/OEBPS/Images/4_02.jpg?sign=1734436936-PbeD5PlMGC7DlLnoHMVs9UgcAPSPmAOR-0-38fd71f73350d12e390001072f72c665)
图6.2 关于TrAdaBoost算法思想的一个直观示例
![](https://epubservercos.yuewen.com/A97C42/27167652804832906/epubprivate/OEBPS/Images/5_01.jpg?sign=1734436936-0sk8frOedzEjA7nLLhBjHhbD4aYjXONa-0-8baf97b549637c1e6e1201902b4fa449)
图6.10 基于锚点的集成学习示意图[100]
![](https://epubservercos.yuewen.com/A97C42/27167652804832906/epubprivate/OEBPS/Images/5_02.jpg?sign=1734436936-KwjriwxFNdkB6JskFbjLsSfzx8arSIJC-0-7372c6e10ffdc1e99166fa8fa5a1991e)
图8.9 拆分架构[130]
![](https://epubservercos.yuewen.com/A97C42/27167652804832906/epubprivate/OEBPS/Images/6_01.jpg?sign=1734436936-g3lo4oEojVn2dWUr4htcJXvRLjnONuQh-0-451963bf49685bb7670ececcd75792a2)
图9.4 视图不足假设[136]
![](https://epubservercos.yuewen.com/A97C42/27167652804832906/epubprivate/OEBPS/Images/6_02.jpg?sign=1734436936-yqSsfK3Gs7CI40pp2pKlIUvMeART04Qd-0-3029cb26b5367d1a5a477410613def2a)
图10.20 风格迁移示意图[202]