摘要
当数据是多模态时,如果在训练阶段没有足够或完整的目标数据可参与训练,则可能导致训练效果较差甚至失败。为了解决该问题,提出了一个基于潜在的低秩约束的不完整模态迁移学习算法(IMTL)。所提算法通过两方面来解决不完整模态问题:一方面,基于低秩约束子空间框架,引入潜在因素来挖掘目标域中缺失的模态信息,然后借助具有完整模态的辅助数据集,通过跨模态或跨数据集方向的迁移学习来帮助模态或数据集之间的数据对齐;另一方面,利用少量标记目标数据来完成监督信息对齐从而保持目标数据在迁移学习过程中的内在结构。实验结果表明,所提算法较之于传统的迁移学习算法有明显优势;即使对于不完整的目标数据,也可以显著地提高分类性能。
When insufficient or incomplete multi-modality data are available in training,the corresponding classification learning may lead to poor training performance or even failure.In order to tackle with this problem,the transfer learning algorithm called IMTL(incomplete modality transfer learning via latent low-rank constraint)is proposed in this paper.The proposed algorithm addresses the incomplete modality problem in two ways.Firstly,latent factors are introduced into a low-rank constrained subspace framework so as to mine missing modality information on the target domain.With the help of an auxiliary yet complete modality dataset,the proposed crossmodality and cross-dataset transfer learning strategy is used to help align data between modalities or datasets.Secondly,a small amount of labeled target data is used to align the supervision information so as to maintain the internal structure of the target data during the transfer learning.Experimental results show that the proposed algorithm outperforms the previous transfer learning algorithms,and significantly improves the classification performance on the adopted incomplete target datasets.
作者
徐光生
王士同
XU Guangsheng;WANG Shitong(School of Artificial Intelligence and Computer Science,Jiangnan University,Wuxi,Jiangsu 214122,China;Key Laboratory of Media Design and Software Technology of Jiangsu Province,Jiangnan University,Wuxi,Jiangsu 214122,China)
出处
《计算机科学与探索》
CSCD
北大核心
2022年第12期2775-2787,共13页
Journal of Frontiers of Computer Science and Technology
基金
江苏省自然科学基金(BK20191331)。
关键词
迁移学习
不完整模态
潜在的低秩约束
transfer learning
incomplete modality
latent low-rank constraint
