Learning to cluster in order to transfer across domains and tasks
Yen-Chang Hsu, Zhaoyang Lv Georgia Institute of Technology Atlanta, GA 30332, USA {yenchang.hsu,、Zsolt Kira Georgia Tech Research Institute Atlanta, GA 30318, USA
Abstract
This paper introduces a novel method to perform transfer learning across domains and tasks, formulating it as a problem of learning to cluster. The key insight is that, in addition to features, we can transfer similarity information and this is sufficient to learn a similarity function and clustering network to perform both domain adaptation and cross-task transfer learning. We begin by reducing categorical information to pairwise constraints, which only considers whether two instances belong to the same class or not (pairwise semantic similarity). This similarity is category-agnostic and can be learned from data in the source domain using a similarity network. We then present two novel approaches for performing transfer learning using this similarity function. First, for unsupervised domain adaptation, we design a new loss function to regularize classification with a constrained clustering loss, hence learning a clustering network with the transferred similarity metric generating the training inputs. Second, for cross-task learning (i.e., unsupervised clustering with unseen categories), we propose a framework to reconstruct and estimate the number of semantic clusters, again using
中文速览
迁移学习通常只迁移特征表示，但当目标任务的类别与源域完全不同时，这种方式会遇到瓶颈。这篇论文提出用"成对语义相似性"（pairwise semantic similarity）作为迁移的核心知识——不管两张图片属于什么具体类别，只判断它们是否同类——从而把分类问题转化为一个"学会聚类"的问题。具体做法是先在有标注的辅助数据集上训练一个相似性预测网络，再用该网络产生的噪声相似性标签驱动一个可端到端训练的可学习聚类目标函数（Learnable Clustering Objective, LCO），让特征提取和聚类分配同步优化；对于跨域迁移，还可以叠加分类损失和域差异损失进一步提升效果。实验表明，在跨任务聚类（Omniglot、ImageNet）和跨域适应（Office-31、SVHN→MNIST）四个基准上均达到当时最优水平，证明仅靠相似性这一轻量级关系知识就能有效桥接不同任务与不同域，为标注成本高昂或类别动态变化的实际场景提供了一条实用路径。
原文 arXiv:1711.10125;中英对照 + 大白话阅读 https://aha.fim.ai/paper/1711.10125v3