Few-shot learning is a fascinating subfield of machine learning that addresses the challenge of training models to recognize and generalize from a limited amount of data. In recent years, the integration of domain adaptation techniques has emerged as a powerful approach to enhance the effectiveness of few-shot learning. This article explores the concept of empowering few-shot learning through domain adaptation, shedding light on the significance, methods, and promising applications of this synergy.

Unsupervised domain adaptation (UDA) has raised a lot of interests in recent years. However, current UDA methods are still not capable enough in dealing with two issues: 1) the scarcity of labeled data in source domain and 2) the need of a general model that can quickly adapt to solve new UDA tasks. To address this situation, we investigate available but rarely-studied setting called few-shot unsupervised domain adaptation (FS-UDA), in which the data of source domain is few-shot per category and the data of target domain remains unlabeled. To realize effective adaptation for FS-UDA tasks in the same source and target domains, we propose a novel meta learning method namely meta-FUDA, which leverages meta learning to perform task-level transfer and domain-level transfer jointly. Extensive experiments demonstrate the promising performance of our method on multiple benchmark data sets

Hyperspectral images (HSIs) are pivotal in remote sensing, providing rich spectral and spatial information for applications such as agriculture, environmental monitoring, and mineral exploration. Unlike traditional RGB images, HSIs capture data across hundreds of contiguous spectral bands, enabling precise material classification based on unique spectral signatures. However, the high dimensionality and the need for pixel-level labeling pose significant challenges, as manual annotation is labor-intensive and time-consuming. To address this, we leverage Domain Adaptation, a transfer learning technique that facilitates knowledge transfer from a well-labeled source domain to a sparsely labeled or unlabeled target domain. In this paper, we focus on few-shot Domain Adaptation for HSI classification, where only a limited number of labeled target samples are available. Our approach integrates Active Learning to iteratively fine-tune a classifier by selecting the most informative target samples, thus reducing domain shift and enhancing model performance. We utilize a framework that generates domain-invariant embeddings by aligning the spectral and spatial features of HSIs from the source and target domains. This method significantly reduces the need for manual labeling while maintaining high classification accuracy. Our experimental results demonstrate the effectiveness of combining few-shot Domain Adaptation with Active Learning, achieving a remarkable 98 % accuracy in hyperspectral image classification with limited labeled data.

Unsupervised domain adaptation aims to learn a classification model from the source domain with much-supervised information, which is applied to the utterly unsupervised target domain. However, collecting enough labeled source samples is difficult in some scenarios, decreasing the effectiveness of previous approaches substantially. Therefore, a more challenging and applicable problem called few-shot unsupervised domain adaptation is considered in this work, where a classifier trained with only a few source labels needs to show strong generalization on the target domain. The prototype-based self-supervised learning method has presented superior performance improvements in addressing this problem, while the quality of the prototype could be further improved. To mitigate this situation, a novel Prototype-Augmented Contrastive Learning is proposed. A new computation strategy is utilized to rectify the source prototypes, which are then used to improve the target prototypes. To better learn semantic information and align features, both in-domain prototype contrastive learning and cross-domain prototype contrastive learning are performed. Extensive experiments are conducted on three widely used benchmarks: Office, OfficeHome, and DomainNet, achieving accuracy improvement of over 3%, 1%, and 0.5%, respectively, demonstrating the effectiveness of the proposed method

Few-shot Test-Time Domain Adaptation focuses on adapting a model at test time to a specific domain using only a few unlabeled examples, addressing domain shift. Prior methods leverage CLIP's strong out-of-distribution (OOD) abilities by generating domain-specific prompts to guide its generalized, frozen features. However, since downstream datasets are not explicitly seen by CLIP, solely depending on the feature space knowledge is constrained by CLIP's prior knowledge. Notably, when using a less robust backbone like ViT-B/16, performance significantly drops on challenging real-world benchmarks. Departing from the state-of-the-art of inheriting the intrinsic OOD capability of CLIP, this work introduces learning directly on the input space to complement the dataset-specific knowledge for frozen CLIP. Specifically, an independent side branch is attached in parallel with CLIP and enforced to learn exclusive knowledge via revert attention. To better capture the dataset-specific label semantics for downstream adaptation, we propose to enhance the inter-dispersion among text features via greedy text ensemble and refinement. The text and visual features are then progressively fused in a domain-aware manner by a generated domain prompt to adapt toward a specific domain. Extensive experiments show our method's superiority on 5 large-scale benchmarks (WILDS and DomainNet), notably improving over smaller networks like ViT-B/16 with gains of \textbf{+5.1} in F1 for iWildCam and \textbf{+3.1\%} in WC Acc for FMoW.

Cross-domain few-shot learning (FSL) often employs adversarial domain adaptation techniques to address the issue of data distribution discrepancies between the source and target domains. However, forcing the alignment of two distinct domains may lead to distortions in class distribution alignment and result in a decrease in classification performance in hyperspectral image (HSI) analysis. Moreover, existing cross-domain methods are often applied to satellite/airborne HSIs as both the source and target domains. It is rarely explored whether the same cross-domain methods can be applied for cross-applications where the source and target domain data could be both satellite/airborne HSIs with lower spatial resolution and unmanned aerial vehicle (UAV) HSIs with higher spatial resolution. To address these issues, this article proposes a novel domain-adaptive FSL network with dual branches, respectively, aiming at domain fusion and domain separation. The domain fusion branch uses a conditional adversarial network to align the global distributions of the two domains, while the domain separation branch introduces a gate mechanism for discriminative feature learning in each domain to achieve independent category distributions. During the experiment, the proposed method is evaluated by performing cross-transfer learning under the condition that low-spatial-resolution hyperspectral data and high-spatial-resolution hyperspectral data are used as source and target data alternately. The experimental results suggest that the proposed method not only mitigates the negative effects of forced alignment in domain fusion, but also holds potential for cross-domain transfer learning between low- and high-spatial-resolution HSIs.

In real-world applications, the environment in which a machine learning system is deployed tends to change due to many factors, such as sample selection bias, prior probability mismatch, and domain shift. This makes it difficult to reliably generalize deep learning models from the training set to real-world scenarios. In addition, data scarcity frequently arises from a large number of applications where annotating data is expensive or requires specialized expertise. As machine learning applications progress into more complex tasks that require models with magnitudes higher Vapnik–Chervonenkis dimensions, more labeled training data are necessary to maintain the same upper bound for the test error. To this end, there is an ever-increasing need for sample efficient learning systems that can adapt to changing environments. This thesis aims to study the generalization of deep learning models in the presence of distribution mismatch and data scarcity. We first study unsupervised domain adaptation, an emerging field of semi-supervised learning that aims to address domain shift with labeled data in the source domain and unlabeled data in the target domain. We propose implicit class-conditioned domain alignment to address between-domain class distribution shift. A theoretical analysis is provided to justify the proposed method by decomposing the empirical domain divergence into class-aligned and class-misaligned divergence, and we show that class-misaligned divergence is detrimental to domain adaptation. We show that our method offers consistent improvements for different adversarial adaptation algorithms. We also propose two meta-learning methods to bridge the gap between gradient and metric-based methods. The first proposal is Conditional class-Aware Meta-Learning where we introduce a metric space to modulate the image representation of a model, resulting in better separated feature representations. Motivated by the discrepancy of the number of training examples between few-shot and real-world medical datasets, the second proposal is to extend few-shot learning to few-to-medium-shot learning. The proposed Task Adaptive Metric Space uses gradient-based fine-tuning to adjust parameters of the metric space to provide more flexibility to metric-based methods. The method adjusts the metric space to better reflect examples of a new medical classification task

Unsupervised domain adaptation aims to facilitate learning tasks within an unlabeled target domain by leveraging labeled information from the source domain. To address the domain discrepancy, the majority of existing methods focus on learning latent features in both the source and target domains. However, the essential challenge in unsupervised domain adaptation lies in locating labeled information, specifically the true labels of the data, within the target domain. In this paper, we propose a unified prompt-tuning framework that converts cross-domain classification into few-shot learning, effectively integrating diverse cross-domain classification tasks into an iterative few-shot learning paradigm. The framework unifies all predicted pseudo labels, templates, and iterated models into a cohesive prompt-tuning model, exhibiting robust scalability for integrating other tuning modalities and offering high availability to minimize the necessity for extensive fine-tuning. Through extensive experiments on several well-known benchmark datasets, we validate the superior performance of our framework compared to other domain adaptation and fine-tuning PLMs methods, which achieves up to an impressive 99% predicting accuracy of true labels within the target domain.

Cross-domain few-shot learning (CD-FSL) requires models to generalize from limited labeled samples under significant distribution shifts. While recent methods enhance adaptability through lightweight task-specific modules, they operate solely in the spatial domain and overlook frequency-specific variations that are often critical for robust transfer. We observe that spatially similar images across domains can differ substantially in their spectral representations, with low and high frequencies capturing complementary semantic information at coarse and fine levels. This indicates that uniform spatial adaptation may overlook these spectral distinctions, thus constraining generalization. To address this, we introduce Frequency Adaptation and Diversion (FAD), a frequency-aware framework that explicitly models and modulates spectral components. At its core is the Frequency Diversion Adapter, which transforms intermediate features into the frequency domain using the discrete Fourier transform (DFT), partitions them into low, mid, and high-frequency bands via radial masks, and reconstructs each band using inverse DFT (IDFT). Each frequency band is then adapted using a dedicated convolutional branch with a kernel size tailored to its spectral scale, enabling targeted and disentangled adaptation across frequencies. Extensive experiments on the Meta-Dataset benchmark demonstrate that FAD consistently outperforms state-of-the-art methods on both seen and unseen domains, validating the utility of frequency-domain representations and band-wise adaptation for improving generalization in CD-FSL.

Abstract Machine learning degrades under distribution shifts, especially with scarce labels. Cross-domain few-shot learning (CDFSL) aims to generalize to novel classes in unseen domains. This dissertation develops a unified theoretical framework and new recognition methods for CDFSL. We extend Empirical Risk Minimization to a two-stage optimization that separates intra- and inter-domain challenges, and organize existing methods into four families: \(\mathcal{D}\)-Extension (instance-level augmentation), \(\Delta\)-Adaptation (distribution alignment), \(\mathcal{H}\)-Constraint (hypothesis-space restriction), and hybrids. For \(\mathcal{D}\)-Extension, we propose Inter-Source Stylization Network (ISSNet), which enriches target instances with styles drawn from unlabeled auxiliary domains, improving generalization via style-driven augmentation. For Δ-Adaptation, we define source-free CDFSL (SF-CDFSL), where source data are inaccessible, and introduce Information-Maximized Distance-aware Contrastive Learning (IM-DCL) to align a pretrained source model with unlabeled target data using distance-aware contrastive learning and information maximization. For \(\mathcal{H}\)-Constraint, we present Step-wise Style Prompt Tuning (StepSPT), which stabilizes learning by constraining the hypothesis space with learnable prompts and step-wise alignment, enabling efficient use of large pretrained models under low-data, source-free conditions. Overall, the work provides a unified perspective and three complementary methods that enhance robustness and generalization in data-constrained cross-domain scenarios, supporting applications in medical imaging, autonomous systems, and remote sensing. Original papers Xu, H., Zhi, S., Sun, S., Patel, V., & Liu, L. (2025). Deep learning for cross-domain few-shot visual recognition: A survey. ACM Computing Surveys, 57(8), 1–37. https://doi.org/10.1145/3718362 https://doi.org/10.1145/3718362 Self-archived version Xu, H., Zhi, S., & Liu, L. (2023). Cross-domain few-shot classification via inter-source stylization. 2023 IEEE International Conference on Image Processing (ICIP), 565–569. https://doi.org/10.1109/ICIP49359.2023.10222701 https://doi.org/10.1109/ICIP49359.2023.10222701 Self-archived version Xu, H., Liu, L., Zhi, S., Fu, S., Su, Z., Cheng, M.-M., & Liu, Y. (2024). Enhancing information maximization with distance-aware contrastive learning for source-free cross-domain few-shot learning. IEEE Transactions on Image Processing, 33, 2058–2073. https://doi.org/10.1109/TIP.2024.3374222 https://doi.org/10.1109/TIP.2024.3374222 Self-archived version Xu, H., Liu, L., Liu, T., Zhi, S., Sun, S., & Cheng, M.-M. (2025). Step-wise distribution-aligned style prompt tuning for source-free cross-domain few-shot learning. IEEE Transactions on Pattern Analysis and Machine Intelligence. Advance online publication. https://doi.org/10.1109/TPAMI.2025.3610039 https://doi.org/10.1109/TPAMI.2025.3610039 Tiivistelmä Perinteiset koneoppimismenetelmät heikkenevät, kun tietojakaumat muuttuvat, erityisesti harvojen näytteiden tilanteissa. Harvojen näytteiden oppiminen yli verkkotunnusten rajojen (cross-domain few-shot learning, CDFSL) pyrkii ratkaisemaan tämän mahdollistamalla mallien yleistymisen uusille luokille ennen näkemättömissä verkkotunnuksissa. Tässä väitöskirjassa tutkitaan CDFSL:ää esittämällä yhtenäinen teoreettinen viitekehys ja kehittämällä uusia menetelmiä tunnistustehtäviin. Aluksi laajennetaan empiirisen riskin minimointi (Empirical Risk Minimization, ERM) kaksivaiheiseksi optimointikehykseksi, joka huomioi sekä verkkotunnusten sisäiset että väliset haasteet. Tämän mallin perusteella nykyiset CDFSL-menetelmät luokitellaan neljään päätyyppiin: D-laajennus (näytetason augmentointi), ∆-sopeutus (jakaumien yhteensovittaminen), H-rajoite (hypoteesitilan rajoittaminen) ja hybridimenetelmät. Luokittelu luo perustan menetelmien systemaattiselle analyysille ja jatkokehitykselle. D-laajennuksen osalta esitetään Inter-Source Stylization Network (ISSNet), joka parantaa kohdeverkkotunnuksen yleistymistä rikastamalla sen näytteitä eri tyylien avulla. ISSNet hyödyntää useita merkitsemättömiä apuverkkotunnuksia ja osoittaa, että tyylipohjainen augmentointi lieventää tehokkaasti verkkotunnusten eroja. ∆-sopeutuksessa määritellään uusi ongelmanasettelu: lähdedomainista riippumaton harvojen näytteiden oppiminen (source-free CDFSL, SF-CDFSL), jossa lähdedata ei ole saatavilla. Tätä varten kehitetään Information-Maximized Distance-aware Contrastive Learning (IM-DCL), joka yhdistää etäisyysherkän kontrastisen oppimisen informaation maksimointiin ja mahdollistaa tehokkaan verkkotunnusten sovituksen ilman lähdedataa. Lisäksi H-rajoitteeseen perustuva Step-wise Style Prompt Tuning (StepSPT) rajoittaa hypoteesitilaa ja parantaa oppimisen vakautta. Menetelmä säätää syöttödatan tyylejä oppivien prompttien avulla ja hienosäätää mallia vaiheittain, mikä vähentää ylioppimisen riskiä ja mahdollistaa laajojen esikoulutettujen mallien hyödyntämisen harvoissa ja lähdedatasta riippumattomissa tilanteissa. Väitöskirja edistää merkittävästi CDFSL-tutkimusta tarjoamalla yhtenäisen teoreettisen kehyksen ja uusia menetelmiä, jotka parantavat mallien robustiutta ja yleistämiskykyä datarajoitteisissa olosuhteissa sekä tukevat sovelluksia lääketieteessä, autonomiassa ja kaukokartoituksessa. Osajulkaisut Xu, H., Zhi, S., Sun, S., Patel, V., & Liu, L. (2025). Deep learning for cross-domain few-shot visual recognition: A survey. ACM Computing Surveys, 57(8), 1–37. https://doi.org/10.1145/3718362 https://doi.org/10.1145/3718362 Rinnakkaistallennettu versio Xu, H., Zhi, S., & Liu, L. (2023). Cross-domain few-shot classification via inter-source stylization. 2023 IEEE International Conference on Image Processing (ICIP), 565–569. https://doi.org/10.1109/ICIP49359.2023.10222701 https://doi.org/10.1109/ICIP49359.2023.10222701 Rinnakkaistallennettu versio Xu, H., Liu, L., Zhi, S., Fu, S., Su, Z., Cheng, M.-M., & Liu, Y. (2024). Enhancing information maximization with distance-aware contrastive learning for source-free cross-domain few-shot learning. IEEE Transactions on Image Processing, 33, 2058–2073. https://doi.org/10.1109/TIP.2024.3374222 https://doi.org/10.1109/TIP.2024.3374222 Rinnakkaistallennettu versio Xu, H., Liu, L., Liu, T., Zhi, S., Sun, S., & Cheng, M.-M. (2025). Step-wise distribution-aligned style prompt tuning for source-free cross-domain few-shot learning. IEEE Transactions on Pattern Analysis and Machine Intelligence. Advance online publication. https://doi.org/10.1109/TPAMI.2025.3610039 https://doi.org/10.1109/TPAMI.2025.3610039 Academic dissertation to be presented with the assent of the Doctoral Programme Committee of Information Technology and Electrical Engineering of the University of Oulu for public defence in the Leena Palotie auditorium (101A) of the Faculty of Medicine (Aapistie 5 A), on 7 November 2025, at 12 noonAbstract Machine learning degrades under distribution shifts, especially with scarce labels. Cross-domain few-shot learning (CDFSL) aims to generalize to novel classes in unseen domains. This dissertation develops a unified theoretical framework and new recognition methods for CDFSL. We extend Empirical Risk Minimization to a two-stage optimization that separates intra- and inter-domain challenges, and organize existing methods into four families: \(\mathcal{D}\)-Extension (instance-level augmentation), \(\Delta\)-Adaptation (distribution alignment), \(\mathcal{H}\)-Constraint (hypothesis-space restriction), and hybrids. For \(\mathcal{D}\)-Extension, we propose Inter-Source Stylization Network (ISSNet), which enriches target instances with styles drawn from unlabeled auxiliary domains, improving generalization via style-driven augmentation. For Δ-Adaptation, we define source-free CDFSL (SF-CDFSL), where source data are inaccessible, and introduce Information-Maximized Distance-aware Contrastive Learning (IM-DCL) to align a pretrained source model with unlabeled target data using distance-aware contrastive learning and information maximization. For \(\mathcal{H}\)-Constraint, we present Step-wise Style Prompt Tuning (StepSPT), which stabilizes learning by constraining the hypothesis space with learnable prompts and step-wise alignment, enabling efficient use of large pretrained models under low-data, source-free conditions. Overall, the work provides a unified perspective and three complementary methods that enhance robustness and generalization in data-constrained cross-domain scenarios, supporting applications in medical imaging, autonomous systems, and remote sensing.Tiivistelmä Perinteiset koneoppimismenetelmät heikkenevät, kun tietojakaumat muuttuvat, erityisesti harvojen näytteiden tilanteissa. Harvojen näytteiden oppiminen yli verkkotunnusten rajojen (cross-domain few-shot learning, CDFSL) pyrkii ratkaisemaan tämän mahdollistamalla mallien yleistymisen uusille luokille ennen näkemättömissä verkkotunnuksissa. Tässä väitöskirjassa tutkitaan CDFSL:ää esittämällä yhtenäinen teoreettinen viitekehys ja kehittämällä uusia menetelmiä tunnistustehtäviin. Aluksi laajennetaan empiirisen riskin minimointi (Empirical Risk Minimization, ERM) kaksivaiheiseksi optimointikehykseksi, joka huomioi sekä verkkotunnusten sisäiset että väliset haasteet. Tämän mallin perusteella nykyiset CDFSL-menetelmät luokitellaan neljään päätyyppiin: D-laajennus (näytetason augmentointi), ∆-sopeutus (jakaumien yhteensovittaminen), H-rajoite (hypoteesitilan rajoittaminen) ja hybridimenetelmät. Luokittelu luo perustan menetelmien systemaattiselle analyysille ja jatkokehitykselle. D-laajennuksen osalta esitetään Inter-Source Stylization Network (ISSNet), joka parantaa kohdeverkkotunnuksen yleistymistä rikastamalla sen näytteitä eri tyylien avulla. ISSNet hyödyntää useita merkitsemättömiä apuverkkotunnuksia ja osoittaa, että tyylipohjainen augmentointi lieventää tehokkaasti verkkotunnusten eroja. ∆-sopeutuksessa määritellään uusi ongelmanasettelu: lähdedomainista riippumaton harvojen näytteiden oppiminen (source-free CDFSL, SF-CDFSL), jossa lähdedata ei ole saatavilla. Tätä varten kehitetään Information-Maximized Distance-aware Contrastive Learning (IM-DCL), joka yhdistää etäisyysherkän kontrastisen oppimisen informaation maksimointiin ja mahdollistaa tehokkaan verkkotunnusten sovituksen ilman lähdedataa. Lisäksi H-rajoitteeseen perustuva Step-wise Style Prompt Tuning (StepSPT) rajoittaa hypoteesitilaa ja parantaa oppimisen vakautta. Menetelmä säätää syöttödatan tyylejä oppivien prompttien avulla ja hienosäätää mallia vaiheittain, mikä vähentää ylioppimisen riskiä ja mahdollistaa laajojen esikoulutettujen mallien hyödyntämisen harvoissa ja lähdedatasta riippumattomissa tilanteissa. Väitöskirja edistää merkittävästi CDFSL-tutkimusta tarjoamalla yhtenäisen teoreettisen kehyksen ja uusia menetelmiä, jotka parantavat mallien robustiutta ja yleistämiskykyä datarajoitteisissa olosuhteissa sekä tukevat sovelluksia lääketieteessä, autonomiassa ja kaukokartoituksessa

Unsupervised domain adaptation (UDA) methods have found broad application in fault diagnosis under variable working conditions. However, the performance of these models significantly deteriorates when the samples in the source and target working conditions are limited. To tackle this issue of few-shot fault diagnosis under variable working conditions, a prototypical contrastive-based domain adaptation method (PCDA) is proposed. The proposed method contains three key modules: a feature extractor, a prototypical contrastive-based few-shot classifier and a domain adaptation module. Initially, the feature extractor is used to extract features from limited samples. Subsequently, the prototypical contrastive-based few-shot classifier is used to identify the fault categories. Alongside, a corresponding prototypical contrastive loss function is designed to dynamically estimate the concentration of feature distribution around each prototype and encourage the representations learned from the same category to be closer. Simultaneously, the method is guided to generate meaningful representations in a self-supervised setting, thereby enhancing the representation ability of the model. Furthermore, the domain adaptation module is utilized to bridge the great gap between two working conditions. Ultimately, PCDA is applied to two cross-domain few-shot bearing fault diagnosis tasks. These tasks include working conditions transfer tasks and machine-type transfer tasks to verify the effectiveness of the proposed method.

Existing few-shot learning (FSL) methods make the implicit assumption that the few target class samples are from the same domain as the source class samples. However, in practice, this assumption is often invalid –the target classes could come from a different domain. This poses an additional challenge of domain adaptation (DA) with few training samples. In this paper, the problem of domain-adaptive few-shot learning (DA-FSL) is tackled, which is expected to have wide use in real-world scenarios and requires solving FSL and DA in a unified framework. To this end, we propose a novel domain-adversarial prototypical network (DAPN) model. It is designed to address a specific challenge in DA-FSL: the DA objective means that the source and target data distributions need to be aligned, typically through a shared domain-adaptive feature embedding space; but the FSL objective dictates that the target domain per class distribution must be different from that of any source domain class, meaning aligning the distributions across domains may harm the FSL performance. How to achieve global domain distribution alignment whilst maintaining source/target per-class discriminativeness thus becomes the key. Our solution is to explicitly enhance the source/target per-class separation before domain-adaptive feature embedding learning, to alleviate the negative effect of domain alignment on FSL. Extensive experiments show that our DAPN outperforms the state-of-the-arts

Cross-domain few-shot learning holds great promise in hyperspectral image classification but still faces significant challenges. Traditional conditional adversarial strategies reduce domain-specific features yet often fail to prevent incorrect class alignment, leading to degraded performance in the target domain. This issue is exacerbated under low class-overlap scenarios, where enforcing global feature alignment distorts class structure and induces negative transfer. To do so, this article proposes a joint domain adaptation weight self-learning (JDAWSL) method for hyperspectral few-shot classification. JDAWSL combines a domain discriminator and a domain projector to suppress domain-specific features while promoting the alignment of semantically consistent domain-shared features, preserving intraclass compactness and interclass separability. This design effectively avoids incorrect class alignment during domain adaptation, preserves class structure, and enhances classification performance in the target domain. In addition, JDAWSL introduces a plug-and-play adaptive learner that dynamically adjusts loss weights based on data characteristics, preventing the overemphasis of an unsuitable domain adaptation loss and improving model robustness to complex data distributions. Experiments confirm that JDAWSL can effectively preserve class structure, mitigate negative transfer, and deliver strong performance and generalization with limited labeled samples.

Due to a lack of labeled samples, deep learning methods generally tend to have poor classification performance in practical applications. Few-shot learning (FSL), as an emerging learning paradigm, has been widely utilized in hyperspectral image (HSI) classification with limited labeled samples. However, the existing FSL methods generally ignore the domain shift problem in cross-domain scenes and rarely explore the associations between samples in the source and target domain. To tackle the above issues, a graph-based domain adaptation FSL (GDAFSL) method is proposed for HSI classification with limited training samples, which utilizes the graph method to guide the domain adaptation learning process in a uniformed framework. First, a novel deep residual hybrid attention network (DRHAN) is designed to extract discriminative embedded features efficiently for few-shot HSI classification. Then, a graph-based domain adaptation network (GDAN), which combines graph construction with domain adversarial strategy, is proposed to fully explore the domain correlation between source and target embedded features. By utilizing the fully explored domain correlations to guide the domain adaptation process, a domain invariant feature metric space is learned for few-shot HSI classification. Comprehensive experimental results conducted on three public HSI datasets demonstrate that GDAFSL is superior to the state-of-the-art with a small sample size

Radar signal recognition (RSR) plays a pivotal role in electronic warfare (EW), as accurately classifying radar signals is critical for informing decision-making. Recent advances in deep learning have shown significant potential in improving RSR in domains with ample annotated data. However, these methods fall short in EW scenarios where annotated radio frequency (RF) data are scarce or impractical to obtain. To address these challenges, we introduce a self-supervised learning (SSL) method which utilises masked signal modelling and RF domain adaption to perform few-shot RSR and enhance performance in environments with limited RF samples and annotations. We propose a two-step approach, first pre-training masked autoencoders (MAE) on baseband in-phase and quadrature (I/Q) signals from diverse RF domains, and then transferring the learned representations to the radar domain, where annotated data are scarce. Empirical results show that our lightweight self-supervised ResNet1D model with domain adaptation achieves up to a $17.5 \% \text{im}-{}$ provement in 1 -shot classification accuracy when pre-trained on in-domain signals (i.e., radar signals) and up to a 16.31 % improvement when pre-trained on out-of-domain signals (i.e., comm signals), compared to its baseline without using SSL. We also present reference results for several MAE designs and pre-training strategies, establishing a new benchmark for few-shot radar signal classification.

Few-shot domain adaptation (FSDA) has become a key approach for cross-domain intrusion detection, enabling models to leverage limited labeled target data under distribution shift. While a wide range of adaptation methods have been proposed, their effectiveness often varies significantly across different transfer scenarios, leading to inconsistent performance and limited interpretability. In this work, it is argued that such variability stems from an overlooked factor: transfer difficulty. This study proposes a transfer-difficulty-aware perspective on FSDA and shows that adaptation behavior is fundamentally dependent on cross-domain compatibility rather than solely on intrinsic domain structure. To this end, a distinction is made between intra-domain separability, which characterizes the internal structure of each domain, and transfer difficulty, which captures how well source-derived representations generalize to the target domain. A set of asymmetric transferability metrics is introduced to quantify this phenomenon and accompanied by a systematic analysis across multiple transfer directions. The results reveal that a domain with strong internal separability does not necessarily yield easy transfer, highlighting that intra-domain structure alone is insufficient to explain cross-domain performance. Furthermore, it is shown that different adaptation strategies exhibit distinct behaviors depending on transfer difficulty: target-only few-shot learning is effective in low-difficulty settings, whereas alignment-based approaches become essential in high-difficulty scenarios. These findings explain the inconsistent performance of existing methods and suggest that domain adaptation should be treated as a transfer-dependent problem rather than a uniform strategy. Ultimately, this work offers both theoretical insights and practical guidance for designing robust cross-domain intrusion detection systems.

Hyperspectral image cross-scene classification (HSICC) tasks are confronted with tremendous challenges due to spectral shift phenomena across scenes and the tough work of obtaining labels. Unsupervised domain adaptation has proven its effectiveness in tackling this issue, but it has a fatal limitation of intending to narrow the disparity between source and target domains by utilizing fully labeled source data and unlabeled target data. However, it is costly even to attain labels from source domains in many cases, rendering sufficient labeling as used in prior work impractical. In this work, we investigate an extreme and realistic scenario where unsupervised domain adaptation methods encounter sparsely labeled source data when handling HSICC tasks, namely, few-shot unsupervised domain adaptation. We propose an end-to-end refined bi-directional prototypical contrastive learning (RBPCL) framework for overcoming the HSICC problem with only a few labeled samples in the source domain. RBPCL captures category-level semantic features of hyperspectral data and performs feature alignment through in-domain refined prototypical self-supervised learning and bi-directional cross-domain prototypical contrastive learning, respectively. Furthermore, our framework introduces the class-balanced multicentric dynamic prototype strategy to generate more robust and representative prototypes. To facilitate prototype contrastive learning, we employ a Siamese-style distance metric loss function to aggregate intra-class features while increasing the discrepancy of inter-class features. Finally, extensive experiments and ablation analysis implemented on two public cross-scene data pairs and three pairs of self-collected ultralow-altitude hyperspectral datasets under different illumination conditions verify the effectiveness of our method, which will further enhance the practicality of hyperspectral intelligent sensing technology.

Contrastive self-supervised learning methods learn to map data points such as images into non-parametric representation space without requiring labels. While highly successful, current methods require a large amount of data in the training phase. In situations where the target training set is limited in size, generalization is known to be poor. Pretraining on a large source data set and fine-tuning on the target samples is prone to overfitting in the few-shot regime, where only a small number of target samples are available. Motivated by this, we propose a domain adaption method for self-supervised contrastive learning, termed Few-Max, to address the issue of adaptation to a target distribution under few-shot learning. To quantify the representation quality, we evaluate Few-Max on a range of source and target datasets, including ImageNet, VisDA, and fastMRI, on which Few-Max consistently outperforms other approaches