This paper explores the ability of Graph Neural Networks (GNNs) in learning various forms of information for link prediction, alongside a brief review of existing link prediction methods. Our analysis reveals that GNNs cannot effectively learn structural information related to the number of common neighbors between two nodes, primarily due to the nature of set-based pooling of the neighborhood aggregation scheme. Also, our extensive experiments indicate that trainable node embeddings can improve the performance of GNN-based link prediction models. Importantly, we observe that the denser the graph, the greater such the improvement. We attribute this to the characteristics of node embeddings, where the link state of each link sample could be encoded into the embeddings of nodes that are involved in the neighborhood aggregation of the two nodes in that link sample. In denser graphs, every node could have more opportunities to attend the neighborhood aggregation of other nodes and encode states of more link samples to its embedding, thus learning better node embeddings for link prediction. Lastly, we demonstrate that the insights gained from our research carry important implications in identifying the limitations of existing link prediction methods, which could guide the future development of more robust algorithms.

Anomaly detection in multivariate time series is a critical task across a range of real-world domains, such as industrial automation and the internet of things. These environments are generally monitored by various types of sensors that produce complex, high-dimensional time-series data with intricate cross-sensor dependencies. While existing methods often utilize sequence modeling or graph neural networks to capture global sensor relationships, they typically treat all sensors uniformly—potentially overlooking the benefit of grouping sensors with similar temporal patterns. To this end, we propose a novel framework called Multi-task Learning Anomaly Detection (MLAD), which leverages clustering techniques to group sensors based on their temporal characteristics, and employs a multi-task learning paradigm to jointly capture both shared patterns across all sensors and specialized patterns within each cluster. MLAD consists of four key modules: (1) sensor clustering based on sensors’ time series, (2) representation learning with a cluster-constrained graph neural network, (3) multi-task forecasting with shared and cluster-specific learning layers, and (4) anomaly scoring. Extensive experiments on three public datasets demonstrate that MLAD achieves superior detection performance over state-of-the-art baselines. Ablation studies further validate the effectiveness of the modules of our MLAD. This study highlights the value of incorporating sensor heterogeneity into model design, which contributes to more accurate and robust anomaly detection in sensor-based monitoring systems.

This paper explores the ability of Graph Neural Networks (GNNs) in learning various forms of information for link prediction, alongside a brief review of existing link prediction methods. Our analysis reveals that GNNs cannot effectively learn structural information related to the number of common neighbors between two nodes, primarily due to the nature of set-based pooling of the neighborhood aggregation scheme. Also, our extensive experiments indicate that trainable node embeddings can improve the performance of GNN-based link prediction models. Importantly, we observe that the denser the graph, the greater such the improvement. We attribute this to the characteristics of node embeddings, where the link state of each link sample could be encoded into the embeddings of nodes that are involved in the neighborhood aggregation of the two nodes in that link sample. In denser graphs, every node could have more opportunities to attend the neighborhood aggregation of other nodes and encode states of more link samples to its embedding, thus learning better node embeddings for link prediction. Lastly, we demonstrate that the insights gained from our research carry important implications in identifying the limitations of existing link prediction methods, which could guide the future development of more robust algorithms.

Remote sensing is playing an increasingly important role in agriculture, enhancing capabilities in monitoring, management, and planning. With the growing availability of data resources such as satellites and drones, there is a corresponding demand for advanced data processing, modeling techniques, and prediction explanations. This survey examines the intersection of temporal explainability, remote sensing, and agriculture through a Systematic Literature Review (SLR). We analyze the utilization of temporal data, the methods employed to uncover temporal patterns, and the types of information explained using Explainable Artificial Intelligence (XAI) in agricultural applications. The findings highlight a diverse spectrum of preprocessing techniques, machine learning models, and temporal analysis strategies tailored to agricultural needs, with a growing emphasis on stage-based temporal modeling and stakeholder-centered explainability. Despite significant advancements, challenges remain in aggregating temporal features, adopting advanced sequential architectures, and evaluating explanation quality. Addressing these challenges is essential for building trust and enabling practical, data-driven solutions in agriculture. Future research should prioritize the development of specialized remote sensing techniques and model architectures that align more closely with agricultural practices, incorporating domain expertise and stakeholder feedback to enhance transparency and trust in AI-driven agriculture.

Deep Neural Networks (DNNs) suffer from domain shift when the test dataset follows a distribution different from the training dataset. Domain generalization aims to tackle this issue by learning a model that can generalize to unseen domains. In this paper, we propose a new approach that aims to explicitly remove domain-specific features for domain generalization. Following this approach, we propose a novel framework called Learning and Removing Domain-specific features for Generalization (LRDG) that learns a domain-invariant model by tactically removing domain-specific features from the input images. Specifically, we design a classifier to effectively learn the domain-specific features for each source domain, respectively. We then develop an encoder-decoder network to map each input image into a new image space where the learned domain-specific features are removed. With the images output by the encoder-decoder network, another classifier is designed to learn the domain-invariant features to conduct image classification. Extensive experiments demonstrate that our framework achieves superior performance compared with state-of-the-art methods.