Detecting anomalous nodes in dynamic graphs is critical for applications such as social network analysis, financial risk management, and cybersecurity. Traditional methods often rely on Euclidean assumptions and linear operations to model temporal node behaviors, which can limit their effectiveness due to entangled norm and direction variations in node embeddings. To address these challenges, we propose SBF-Net (Spherical Behavior and Frequency-aware Network), a novel framework that projects node features onto a unit hypersphere to better disentangle norm and directional changes, enabling more accurate behavior quantification. To mitigate the scarcity of anomalous samples, we employ a Deep Belief Network-based anomaly synthesizer to generate diverse synthetic anomalies. Additionally, frequency domain analysis is incorporated to capture subtle and high-frequency anomaly patterns. Extensive experiments on multiple real-world dynamic graph datasets demonstrate that SBF-Net consistently improves anomaly detection performance over several strong baselines. For instance, on the Wikipedia dataset, SBF-Net achieves improvements of 6.32% in AUC and 7.93% in precision over existing methods.

Transformer-based detectors have advanced small-object detection, but they often remain inefficient and vulnerable to background-induced query noise, which motivates deep decoders to refine low-quality queries. We present HELP (Heatmap-guided Embedding Learning Paradigm), a noise-aware positional-semantic fusion framework that studies where to embed positional information by selectively preserving positional encodings in foreground-salient regions while suppressing background clutter. Within HELP, we introduce Heatmap-guided Positional Embedding (HPE) as the core embedding mechanism and visualize it with a heatbar for interpretable diagnosis and fine-tuning. HPE is integrated into both the encoder and decoder: it guides noise-suppressed feature encoding by injecting heatmap-aware positional encoding, and it enables high-quality query retrieval by filtering background-dominant embeddings via a gradient-based mask filter before decoding. To address feature sparsity in complex small targets, we integrate Linear-Snake Convolution to enrich retrieval-relevant representations. The gradient-based heatmap supervision is used during training only, incurring no additional gradient computation at inference. As a result, our design reduces decoder layers from eight to three and achieves a 59.4% parameter reduction (66.3M vs. 163M) while maintaining consistent accuracy gains under a reduced compute budget across benchmarks. Code Repository: https://github.com/yidimopozhibai/Noise-Suppressed-Query-Retrieval.

In dynamic graphs, detecting anomalous nodes faces challenges due to sample imbalance, stemming from the scarcity of anomalous samples and feature representation bias. Existing methods often use unsupervised or semi-supervised learning to extract anomalous samples from unlabeled data, but struggle to obtain enough anomalous instances due to their low occurrence. Moreover, GNN-based approaches often prioritize normal samples, neglecting rare anomalies. To address these issues, we propose the Anomaly Balance Network (ABNet), designed to alleviate sample imbalance and enhance anomaly detection. ABNet includes three key components: a feature extractor that compares node features across time points to avoid bias, an anomaly augmenter that amplifies anomaly details and generates diverse anomalous samples, and an anomaly detector using meta-learning to adapt to graph evolution. Experimental results show that ABNet outperforms existing methods on three real-world datasets, effectively addressing sample imbalance.

Few-shot anomalous node detection in dynamic networks has been extensively investigated in the field of research. In this few-shot scenario, the detection of these anomalous nodes is particularly challenging due to the continuously evolving network topology and data distribution over time, which is known as concept drift. Concept drift refers to the phenomenon where the underlying concepts or patterns in the data generation process change over time, leading to varying data distributions across different periods. Due to these changes in data distribution, the patterns learned during training may become invalid under the new data distribution. Existing models primarily aim to enhance the representation of evolving node attributes and relationships to mitigate the impact of concept drift in few-shot scenarios. However, the scarcity of anomalous samples further limits the model's ability to learn new patterns, thereby reducing its effectiveness in addressing concept drift in few-shot scenarios. To address this challenge, we propose the multitask asynchronous metalearning framework (MAMF), which aims to mitigate bias induced by concept drift in few-shot anomalous node detection. Our framework consists of four main components: a feature extractor, an anomaly simulator, an asynchronous learner, and a type detector. The feature extractor captures the relative variations of each node in an evolving graph stream. The anomaly simulator uses generative adversarial models to learn anomaly distributions and generate samples at different time intervals. The asynchronous learner samples from various time distributions to create metatasks for anomalous node detection, allowing it to adapt to changes between these distributions. To aid in few-shot anomalous node detection, the type detector is used for anomaly type recognition. Our framework achieves AUC improvements of 5.12%, 6.87%, and 1.91% over the best existing methods on Wikipedia, Reddit, and Mooc datasets, respectively, demonstrating its effectiveness and robustness in adapting to concept drift and detecting anomalous nodes.

With the growing popularity of artificial intelligence in drug discovery, many deep-learning technologies have been used to automatically predict unknown drug-target interactions (DTIs). A unique challenge in using these technologies to predict DTI is fully exploiting the knowledge diversity across different interaction types, such as drug-drug, drug-target, drug-enzyme, drug-path, and drug-structure types. Unfortunately, existing methods tend to learn the specifical knowledge on each interaction type and they usually ignore the knowledge diversity across different interaction types. Therefore, we propose a multitype perception method (MPM) for DTI prediction by exploiting knowledge diversity across different link types. The method consists of two main components: a type perceptor and a multitype predictor. The type perceptor learns distinguished edge representations by retaining the specifical features across different interaction types; this maximizes the prediction performance for each interaction type. The multitype predictor calculates the type similarity between the type perceptor and predicted interactions, and the domain gate module is reconstructed to assign an adaptive weight to each type perceptor. Extensive experiments demonstrate that our proposed MPM outperforms the state-of-the-art methods in DTI prediction.