Domain adaptation addresses the common problem when the target distribution generating our test data drifts from the source (training) distribution. While absent assumptions, domain adaptation is impossible, strict conditions, e.g. covariate or label shift, enable principled algorithms. Recently-proposed domain-adversarial approaches consist of aligning source and target encodings, often motivating this approach as minimizing two (of three) terms in a theoretical bound on target error. Unfortunately, this minimization can cause arbitrary increases in the third term, e.g. they can break down under shifting label distributions. We propose asymmetrically-relaxed distribution alignment, a new approach that overcomes some limitations of standard domain-adversarial algorithms. Moreover, we characterize precise assumptions under which our algorithm is theoretically principled and demonstrate empirical benefits on both synthetic and real datasets.
Recent advances in domain adaptation establish that requiring a low risk on the source domain and equal feature marginals degrade the adaptation's performance. At the same time, empirical evidence shows that incorporating an unsupervised target domain term that pushes decision boundaries away from the high-density regions, along with relaxed alignment, improves adaptation. In this paper, we theoretically justify such observations via a new bound on the target risk, and we connect two notions of relaxation for divergence, namely $β-$relaxed divergences and localization. This connection allows u…
Univ Lyon（里昂大学）；Ecole Centrale de Lyon（里昂中央理工学院）；CNRS（法国国家科学研究中心）；INSA Lyon（里昂国立应用科学学院）；Univ Claude Bernard Lyon 1（里昂第一大学克洛德·贝尔纳）；Univ Louis Lumière Lyon 2（里昂第二大学路易·卢米埃尔）；Thales LAS France（泰雷兹LAS法国公司）
Deep neural networks often fail to generalize outside of their training distribution, in particular when only a single data domain is available during training. While test-time adaptation has yielded encouraging results in this setting, we argue that, to reach further improvements, these approaches should be combined with training procedure modifications aiming to learn a more diverse set of patterns. Indeed, test-time adaptation methods usually have to rely on a limited representation because of the shortcut learning phenomenon: only a subset of the available predictive patterns is learned wi…
Chinese University of Hong Kong, Shen Zhen（香港中文大学（深圳））；The Hong Kong University of Science and Technology（香港科技大学）；Huawei Noah's Ark Lab（华为诺亚方舟实验室）；Tsinghua University（清华大学）
Semi-supervised domain adaptation (SSDA), which aims to learn models in a partially labeled target domain with the assistance of the fully labeled source domain, attracts increasing attention in recent years. To explicitly leverage the labeled data in both domains, we naturally introduce a conditional GAN framework to transfer images without changing the semantics in SSDA. However, we identify a label-domination problem in such an approach. In fact, the generator tends to overlook the input source image and only memorizes prototypes of each class, which results in unsatisfactory adaptation per…
Massachusetts General Hospital（麻省总医院）；Harvard Medical School（哈佛医学院）
Unsupervised domain adaptation (UDA) has been a vital protocol for migrating information learned from a labeled source domain to facilitate the implementation in an unlabeled heterogeneous target domain. Although UDA is typically jointly trained on data from both domains, accessing the labeled source domain data is often restricted, due to concerns over patient data privacy or intellectual property. To sidestep this, we propose "off-the-shelf (OS)" UDA (OSUDA), aimed at image segmentation, by adapting an OS segmentor trained in a source domain to a target domain, in the absence of source domai…
Unsupervised Domain Adaptation (UDA) has attracted a lot of attention in the last ten years. The emergence of Domain Invariant Representations (IR) has improved drastically the transferability of representations from a labelled source domain to a new and unlabelled target domain. However, a potential pitfall of this approach, namely the presence of \textit{label shift}, has been brought to light. Some works address this issue with a relaxed version of domain invariance obtained by weighting samples, a strategy often referred to as Importance Sampling. From our point of view, the theoretical as…
University of Illinois at Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）
The vast majority of existing algorithms for unsupervised domain adaptation (UDA) focus on adapting from a labeled source domain to an unlabeled target domain directly in a one-off way. Gradual domain adaptation (GDA), on the other hand, assumes a path of $(T-1)$ unlabeled intermediate domains bridging the source and target, and aims to provide better generalization in the target domain by leveraging the intermediate ones. Under certain assumptions, Kumar et al. (2020) proposed a simple algorithm, Gradual Self-Training, along with a generalization bound in the order of $e^{O(T)} \left(\varepsi…
On the Transferability and Discriminability of Repersentation Learning in Unsupervised Domain Adaptation
National Key Laboratory of Space Integrated Information System, Institute of Software Chinese Academy of Sciences（空间信息集成信息系统国家重点实验室，软件研究所）；University of Chinese Academy of Sciences（中国科学院大学）；National Key Laboratory of Space Integrated Information System, Department of Computer Science and Technology, Tsinghua University（空间信息集成信息系统国家重点实验室，清华大学计算机科学与技术系）；Artificial Intelligence Thrust, Information Hub, Department of Computer Science & Engineering, School of Engineering, Hong Kong University of Science and Technology（人工智能 thrust，信息枢纽，工程学院计算机科学与工程系，香港科学与技术大学）
ReLoRA: Knowledge-Reusing Adaptation for Fast Rollout of Evolving LLM Services
School of Computer Science and Technology, University of Science and Technology of China（计算机科学与技术学院，中国科学技术大学）；Suzhou Institute for Advanced Research, University of Science and Technology of China（苏州先进研究院，中国科学技术大学）
Offline imitation learning (IL) promises the ability to learn performant policies from pre-collected demonstrations without interactions with the environment. However, imitating behaviors fully offline typically requires numerous expert data. To tackle this issue, we study the setting where we have limited expert data and supplementary suboptimal data. In this case, a well-known issue is the distribution shift between the learned policy and the behavior policy that collects the offline data. Prior works mitigate this issue by regularizing the KL divergence between the stationary state-action d…
Massachusetts General Hospital（麻省总医院）；Harvard Medical School（哈佛医学院）；Facebook Artificial Intelligence（脸书人工智能研究院）
Unsupervised domain adaptation (UDA) aims to transfer knowledge learned from a labeled source domain to an unlabeled and unseen target domain, which is usually trained on data from both domains. Access to the source domain data at the adaptation stage, however, is often limited, due to data storage or privacy issues. To alleviate this, in this work, we target source free UDA for segmentation, and propose to adapt an ``off-the-shelf" segmentation model pre-trained in the source domain to the target domain, with an adaptive batch-wise normalization statistics adaptation framework. Specifically, …
On Convergence of Adam for Stochastic Optimization under Relaxed Assumptions
Center for Data Science and School of Mathematical Sciences, Zhejiang University（浙江大学数据科学中心与数学科学学院）；Center for Data Science, Zhejiang University（浙江大学数据科学中心）
MIT CSAIL（麻省理工学院计算机科学与人工智能实验室）；MIT-IBM Watson AI Lab（麻省理工学院-IBM沃森人工智能实验室）；UC Santa Barbara（加州大学圣巴巴拉分校）
We study the problem of aligning the supports of distributions. Compared to the existing work on distribution alignment, support alignment does not require the densities to be matched. We propose symmetric support difference as a divergence measure to quantify the mismatch between supports. We show that select discriminators (e.g. discriminator trained for Jensen-Shannon divergence) are able to map support differences as support differences in their one-dimensional output space. Following this result, our method aligns supports by minimizing a symmetrized relaxed optimal transport cost in the …
