Learning Transferable Features with Deep Adaptation Networks
Mingsheng Long†♯ Yue Cao† Jianmin Wang† Michael I. Jordan♯ †School of Software, TNList Lab for Info. Sci.、Tech., Institute for Data Science, Tsinghua University, China ♯Department of Electrical Engineering and Computer Science, University of California, Berkeley, CA, USA
Abstract
Recent studies reveal that a deep neural network can learn transferable features which generalize well to novel tasks for domain adaptation. However, as deep features eventually transition from general to specific along the network, the feature transferability drops significantly in higher layers with increasing domain discrepancy. Hence, it is important to formally reduce the dataset bias and enhance the transferability in task-specific layers. In this paper, we propose a new Deep Adaptation Network (DAN) architecture, which generalizes deep convolutional neural network to the domain adaptation scenario. In DAN, hidden representations of all task-specific layers are embedded in a reproducing kernel Hilbert space where the mean embeddings of different domain distributions can be explicitly matched. The domain discrepancy is further reduced using an optimal multi-kernel selection method for mean embedding matching. DAN can learn transferable features with statistical guarantees, and can scale linearly by unbiased estimate of kernel embedding. Extensive empirical evidence shows that the proposed architecture yields state-of-the-art image classification error rates on standard domain
中文速览
跨域图像识别长期面临一个难题：在源域（有标注）训练好的深度神经网络，搬到分布不同的目标域（无标注）后性能会大幅下滑，根本原因是网络越靠近输出的高层特征越与具体任务绑定，跨域迁移能力急剧下降。为此，作者提出深度适应网络（Deep Adaptation Network, DAN），核心做法是把网络所有任务相关层（fc6–fc8）的隐层表示同时嵌入再生核希尔伯特空间（RKHS），用多核最大均值差异（MK-MMD）显式对齐源域与目标域的分布，并设计了线性时间的无偏核均值估计，让整个框架在保持统计保证的同时也能高效训练。在标准域自适应基准（Office数据集等）上的大量实验表明，DAN在图像分类错误率上全面超越此前最优方法。这项工作的意义在于，它把分布对齐直接融入深度网络的训练目标，用一个端到端的框架同时解决了"适应哪些层"和"如何高效匹配分布"两个关键问题，为后续深度迁移学习研究奠定了重要基础。
原文 arXiv:1502.02791;中英对照 + 大白话阅读 https://aha.fim.ai/paper/1502.02791v2