.png){kind=icon}
返回
Supervised and semi-supervised domain generalization via consistency learning and maximum coding rate reduction
.png){kind=icon}
PRE
AI
DOI：10.1016/j.neucom.2026.132830.png)
.png){kind=icon}
原文求助.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
Domain generalization (DG) seeks to train a classification model that generalizes well to unseen target domains by learning from multiple source domains without access to target domain data. The source and target domains share related but different joint distributions, each lying on a Riemannian statistical manifold. The effectiveness of classifying mixed-distribution data depends on their separability (or discriminability). A common assumption is that the distribution of each class possesses a low-dimensional intrinsic structure, implying that its support lies on a low-dimensional submanifold. In this article, to partition between classes, we employ maximum coding rate reduction (MCR2), an information-theoretic metric that maximizes the difference in coding rates between the entire dataset and the sum of individual classes. Additionally, we introduce our loss functions to enforce feature consistency between original and augmented versions of data with the same label in the feature space, enhancing the model’s generalization within a unified framework. Our method is applicable to both supervised and semi-supervised domain generalization scenarios, and we demonstrate its effectiveness against baseline approaches on standard benchmark datasets, including PACS, Office-Home, VLCS, and Digits-DG.
期刊
IF:
6.5
论文数:
2.5W
被引数:
6.5W
机构
暂无机构信息
引用论文
Riemannian representation learning for multi-source domain adaptation多源域自适应的Riemannian表示学习.png)
PATTERN RECOGNITION
PATTERN RECOGNITIONIF7.6
PREAI
分享收藏Iterative compression towards in-distribution features in domain generalization在领域泛化中，通过迭代压缩朝向分布内特征NEUROCOMPUTING
NEUROCOMPUTINGIF6.5
PREAI
分享收藏
Class and Domain Low-rank Tensor Learning for Multi-source Domain Adaptation多源域适应的类与域低秩张量学习PATTERN RECOGNITION
PATTERN RECOGNITIONIF7.6
PREAI
分享收藏Open-set domain generalization for fault diagnosis through data augmentation and a dual-level weighted mechanism通过数据增强和双层加权机制进行故障诊断的开集域泛化
PREAI
分享收藏
查看更多内容