Distribution shifts on graphs—the discrepancies in data distribution between training and employing a graph machine learning model—are ubiquitous and often unavoidable in real-world applications. These shifts may severely deteriorate model performance, posing significant challenges for reliable graph machine learning. In recent years, there has been a surge in research on graph machine learning specifically designed to tackle such distribution shifts, aiming to train models to achieve satisfactory performance on Out-of-Distribution (OOD) test data. This survey provides an up-to-date and forward-looking review of deep graph learning under distribution shifts. We categorize the field into three primary scenarios: graph OOD generalization, training-time graph OOD adaptation, and test-time graph OOD adaptation. We begin by formally formulating the problems and discussing various types of distribution shifts that can affect graph learning, such as covariate shifts and concept shifts. To provide a structured understanding of the literature, we introduce a systematic taxonomy that classifies existing methods into model-centric and data-centric approaches, investigating the techniques used in each category. We also summarize commonly used datasets in this research area to facilitate further investigation. Finally, we point out promising research directions and the corresponding challenges to encourage further study in this vital domain. Additionally, we provide a continuously updated reading list at https://github.com/kaize0409/Awesome-Graph-OOD .

Intracerebral hemorrhage (ICH) often leads to high disability and mortality rates, and accurate and rapid segmentation of hematoma regions is crucial for condition assessment and treatment planning. Existing neural network methods are unsatisfactory due to the irregularity of hematoma morphology and location, and limited generalization performance. To address these issues, this paper proposes an improved U-Net architecture, which incorporates two innovative modules: an Adaptive High-Frequency Domain Enhancement module and a Context-Aware Similarity Fusion module. The former leverages ICH's highsignal characteristics through frequency-domain enhancement and intensity-guided weighting to selectively enhance ICH regions, thereby improving the network's sensitivity to ICH. The latter enables efficient integration of multi-level contextual information by calculating feature similarity between decoder levels and adaptively filtering and fusing features according to spatial consistency, thereby improving segmentation performance. Experimental results demonstrate that our method achieves improved performance across multiple evaluation metrics compared to existing methods, and provides a more reliable tool for automated diagnosis of ICH.

Few-shot Semantic Segmentation (FSS) aims to adapt a pretrained model to unseen classes using only a handful of labeled examples. Although recent approaches built upon the Segmentation Anything Model (SAM) have shown promising progress, most of them fail to sufficiently incorporate intervoxel spatial information and explicit boundary modeling, which leads to inaccurate or inconsistent 3D predictions. This reduces such models effectiveness in 3D medical segmentation, especially under few-shot annotation available. To mitigate these issues, we propose DBNet, a boundary-aware dual-branch architecture tailored for few-shot 3D medical image segmentation. DBNet can delineate 3D anntomical structures from only one or a few annotated volumes by integrating prior knowledge into convolutional operations and preserving volumetric coherence through the exploitation of inter-voxel spatial cues. Specifically, to mine global contextual features alongside capturing finegrained local boundary details, DBNet embeds prior knowledge into convolution operations for explicit boundary modeling and employs a SAM-based encoder with scale and shift factors (SSF) to extract global semantic representations. Moreover, to compensate the inter-slice spatial information lost in the 2dpretrained SAM encoder, a voxel information reconstruction strategy is introduced to recover inter-slice dependencies, resulting in strengthening volumetric context modeling. Experimental results on two challenging medical benchmarks demonstrate that DBNet exceeds the performance of earlier state-of-the-art 3D segmentation methods under few-shot annotation conditions.

Inspired by the success of transformers in natural language processing, vision transformers have been proposed to address a wide range of computer vision tasks, such as image classification, object detection and image segmentation, and they have achieved very promising performance. However, the robustness of vision transformers has been relatively under-explored. Recent studies have revealed that pre-trained vision transformers are also vulnerable to white-box adversarial attacks on the downstream image classification tasks. The adversarial attacks (e.g., FGSM and PGD) designed for convolutional neural networks (CNNs) can also cause severe performance drop for vision transformers. In this paper, we evaluate the robustness of vision transformers fine-tuned with the off-the-shelf methods under adversarial attacks on CIFAR-10 and CIFAR-100 and further propose a data-augmented virtual adversarial training approach called MixVAT, which is able to enhance the robustness of pre-trained vision transformers against adversarial attacks on the downstream tasks with the unlabelled data. Extensive results on multiple datasets demonstrate the superiority of our approach over baselines on adversarial robustness, without compromising generalization ability of the model.

Unsupervised domain adaptation (UDA) has become an appealing approach for knowledge transfer from a labeled source domain to an unlabeled target domain. However, when the classes in source and target domains are imbalanced, most existing UDA methods experience significant performance drop, as the decision boundary usually favors the majority classes. Some recent class-imbalanced domain adaptation (CDA) methods aim to tackle the challenge of biased label distribution by exploiting pseudo-labeled target samples during the training process. However, these methods suffer from the issues with unreliable pseudo labels and error accumulation during training. In this paper, we propose a pairwise adversarial training approach for class-imbalanced domain adaptation. Unlike conventional adversarial training in which the adversarial samples are obtained from the$\ell _{p}$ball of the original samples, we generate adversarial samples from the interpolated line of the aligned pairwise samples from source and target domains. The pairwise adversarial training (PAT) is a novel data-augmentation method which can be integrated into existing unsupervised domain adaptation (UDA) models to tackle the CDA problem. Inspired by the noise injection, we also extend the pairwise adversarial training to noisy pairwise adversarial training (nPAT), in which the random noise is injected into the generation of the adversarial samples. In our study, we evaluate our proposed methods as well as the baselines on three major benchmark datasets, namely Office-Home, DomainNet and Office-31. For Office-Home and Office-31, we sample the data according to the Reversely-unbalanced Source and Unbalanced Target (RS-UT) protocol so that the class distribution can be imbalanced. The extensive experimental results show that UDA models integrated with our proposed nPAT can achieve prominent improvements on most tasks compared to the baseline methods as well as the state-of-the-art CDA methods. The average accuracy of our nPAT can achieve 66.56% and 80.22% on Office-Home and DomainNet, respectively, which are higher than that of the second-best methods. Besides, Experiments also show that our method is robust to the unreliability of the pseudo labels.

Inspired by the success of transformers in natural language processing, vision transformers have been proposed to address a wide range of computer vision tasks, such as image classification, object detection and image segmentation, and they have achieved very promising performance. However, the robustness of vision transformers has been relatively under-explored. Recent studies have revealed that pre-trained vision transformers are also vulnerable to white-box adversarial attacks on the downstream image classification task. The adversarial attacks (e.g., FGSM and PGD) designed for convolutional neural networks (CNNs) can also cause severe performance drop for vision transformers. In this paper, we evaluate the robustness of vision transformers fine-tuned with the off-the-shelf methods under adversarial attacks on CIFAR-10 and CIFAR-100. We further propose a data-augmented virtual adversarial training approach called MixVAT, which is able to enhance the robustness of pre-trained vision transformers against adversarial attacks on the downstream tasks with the unlabelled data. Extensive results on multiple datasets demonstrate the superiority of our approach over baselines on adversarial robustness, without compromising generalization ability of the model.

Brook trout (Salvelinus fontinalis) is a freshwater fish species of ecological, economic, and cultural importance in eastern North America. Estimating the abundance, movement, and survival of brook trout in the wild is an important task for environmental management, and current methods often involve physical tagging or collection of DNA samples for each of the fish as their unique identifier. However, this process is expensive and inefficient. Meanwhile, although deep learning methods have proven effective for individual recognition of humans, it remains challenging to apply this system to wildlife biology due to fewer available images, different biometric patterns, and relatively poor image quality. In this paper, we develop a framework to automate the process of individual recognition of brook trout. Distinguished from simply adopting traditional feature descriptors (e.g., SIFT and HOG) or using deep neural networks on the raw images, our framework utilizes multiple modalities consisting of the region of interest and gray-scaled pigmentation patterns. We use these multiple modality features in a Convolutional Neural network to generate feature vectors as fish descriptors. These descriptors are then used to distinguish individual brook trout by ranking their relative distance in latent space. Our experimental framework demonstrates better results than baseline methods such as SIFT and HOG while being more robust to distortions characteristic of large imagery datasets collected through crowdsourcing and citizen science.

Unsupervised domain adaptation (UDA) has become an appealing approach for knowledge transfer from a labeled source domain to an unlabeled target domain. However, when the classes in source and target domains are imbalanced, most existing UDA methods experience significant performance drop, as the decision boundary usually favors the majority classes. Some recent class-imbalanced domain adaptation (CDA) methods aim to tackle the challenge of biased label distribution by exploiting pseudo-labeled target samples during training process. However, these methods suffer from the issues with unreliable pseudo labels and error accumulation during training. In this paper, we propose a pairwise adversarial training approach for class-imbalanced domain adaptation. Unlike conventional adversarial training in which the adversarial samples are obtained from the lp ball of the original samples, we generate adversarial samples from the interpolated line of the aligned pairwise samples from source and target domains. The pairwise adversarial training (PAT) is a novel data-augmentation method which can be integrated into existing UDA models to tackle with the CDA problem. Experimental results and ablation studies show that the UDA models integrated with our method achieve considerable improvements on benchmarks compared with the original models as well as the state-of-the-art CDA methods. Our source code is available at: https://github.com/DamoSWL/Pairwise-Adversarial-Training

Numerous technologies now exist for promoting more active lifestyles. However, while quantitative data representations (e.g., charts, graphs, and statistical reports) typify most health tools, growing evidence suggests such feedback can not only fail to motivate behavior but may also harm self-integrity and fuel negative mindsets about exercise. Our research seeks to devise alternative, more qualitative schemes for encoding personal information. In particular, this paper explores the design of data-driven narratives, given the intuitive and persuasive power of stories. We present WhoIsZuki, a smartphone application that visualizes physical activities and goals as components of a multi-chapter quest, where the main character's progress is tied to the user's. We report on our design process involving online surveys, in-lab studies, and in-the-wild deployments, aimed at refining the interface and the narrative and gaining a deep understanding of people's experiences with this type of feedback. From these insights, we contribute recommendations to guide future development of narrative-based applications for motivating healthy behavior.

Graph cuts is an image segmentation method by which the region and boundary information of objects can be revolved comprehensively. Because of the complex spatial characteristics of high-dimensional images, time complexity and segmentation accuracy of graph cuts methods for high-dimensional images need to be improved. This paper proposes a new three-dimensional multilevel banded graph cuts model to increase its accuracy and reduce its complexity. Firstly, three-dimensional image is viewed as a high-dimensional space to construct three-dimensional network graphs. A pyramid image sequence is created by Gaussian pyramid downsampling procedure. Then, a new energy function is built according to the spatial characteristics of the three-dimensional image, in which the adjacent points are expressed by using a 26-connected system. At last, the banded graph is constructed on a narrow band around the object/background. The graph cuts method is performed on the banded graph layer by layer to obtain the object region sequentially. In order to verify the proposed method, we have performed an experiment on a set of three-dimensional colon CT images, and compared the results with local region active contour and Chan–Vese model. The experimental results demonstrate that the proposed method can segment colon tissues from three-dimensional abdominal CT images accurately. The segmentation accuracy can be increased to 95.1% and the time complexity is reduced by about 30% of the other two methods.

The long propagation delay and low data rate of acoustic communication make the exchange of control messages time-consuming. Therefore, traditional handshaking-based Medium Access Control (MAC) protocols do not perform well in Underwater Acoustic Sensor Networks (UASN). In this paper, we propose Stentorian MAC for delay-tolerant UASN with heavy data load. Instead of handshaking, Stentorian MAC uses one-way declaration of transmission plans which can reach all contention peers of the sender. With slotted timing and separation between channel contention and data transmission, it employs a distributed and asynchronous planning strategy, which not only eliminates data collision, but also arranges as many concurrent transmissions as possible, in order to achieve high throughput. We evaluated the concurrency, throughput and packet delay of Stentorian MAC through simulation of random transmission and data aggregation, and confirmed that it has met our design goal.