Multi-source Few-shot Domain Adaptation

Abstract

Multi-source Domain Adaptation (MDA) aims to transfer predictive models from multiple, fully-labeled source domains to an unlabeled target domain. However, in many applications, relevant labeled source datasets may not be available, and collecting source labels can be as expensive as labeling the target data itself. In this paper, we investigate Multi-source Few-shot Domain Adaptation (MFDA): a new domain adaptation scenario with limited multi-source labels and unlabeled target data. As we show, existing methods often fail to learn discriminative features for both source and target domains in the MFDA setting. Therefore, we propose a novel framework, termed Multi-Source Few-shot Adaptation Network (MSFAN), which can be trained end-to-end in a non-adversarial manner. MSFAN operates by first using a type of prototypical, multi-domain, self-supervised learning to learn features that are not only domain-invariant but also class-discriminative. Second, MSFAN uses a small, labeled support set to enforce feature consistency and domain invariance across domains. Finally, prototypes from multiple sources are leveraged to learn better classifiers. Compared with state-of-the-art MDA methods, MSFAN improves the mean classification accuracy over different domain pairs on MFDA by 20.2%, 9.4%, and 16.2% on Office, Office-Home, and DomainNet, respectively. * Equal contribution; work done when Zangwei was in Nanjing University.Preprint. Under review.

SunCast: Solar Irradiance Nowcasting from Geosynchronous Satellite Data

“…Future work may involve including more number of input images, adding channels (e.g. infrared, near-infrared and visible) to these input images to see how this may improve performance, as well as adapting self-supervised and domain adaptation methods to this work [24,25,26]. In addition, NOAA operates geostationary satellites that collect similar spectral data all across the globe.…”

Augmenting and Aligning Snippets for Few-Shot Video Domain Adaptation

“…AcroFOD [11] explores FSDA for object detection by applying multi-level spatial augmentation and filtering target-irrelevant source data. There are also works as in [68,40,38,65] that combine domain adaptation (DA) with few-shot learning (FSL), yet we differ them in the assumption of similar target and source classes and only limited target data accessible, which is more realistic. More recently, there have been a few early research on FSVDA, including PASTN [12] that constructs pairwise adversarial networks performed across source-target video pairs, while PTC [13] further leverages optical flow features.…”