Rapid development of web services has led to an explosion of multimodal content, making multimodal recommender systems (MRSs) vital tools for mitigating information overload. Current MRSs have achieved remarkable progress by incorporating advanced technologies such as Graph Neural Networks (GNNs) and Large Language Models (LLMs). However, these studies still suffer from the semantic shift problem. Generally, item's multimodal content usually contain multiple objects, including target object (core content of item) and auxiliary objects (decorations of item). Existing MRSs overlooked this distinction, failing to prevent auxiliary objects from dominating the representation, leading to biased item representation. To address this issue, we propose a model-agnostic framework ''TargetMR''. Concretely, TargetMR comprises two core modules, including Object Disentangler and Object Identifier. The Object Disentangler decouples item text and image into multiple objects via text syntactic parsing and image segmentation. The Object Identifier performs knowledge distillation based on LLMs to efficiently identify the target text object. It then identifies the target image object through cross-modal semantic evaluation. Moreover, this module refines the representation of image target object by optimizing the semantic correlation. Owing to the model-agnostic design of TargetMR, it can be integrated into various backbone MRSs. Extensive experiments on three benchmark datasets show that TargetMR consistently improves the performance of five backbone MRSs, with an average improvement of 12.26%. Our codes are available at https://github.com/gutang-97/TargetMR/.

Federated recommender systems (FedRecs) can effectively address the problem of information overload without using users' private data. Based on federated learning, FedRecs have achieved significant success in protecting user privacy. FedRecs allow users to train recommendation models locally and upload the trained gradients to the server, which effectively protect privacy. However, this decentralized learning paradigm also makes federated recommender systems vulnerable to poisoning attacks. Attackers can manipulate the recommendation results by poisoning local training datasets or uploading toxic gradients. This paper provides an overview of poisoning attacks against federated recommender systems, introducing them from three perspectives. Classify and discuss based on the purpose of the attacker, the prior knowledge of the attacker, and the method of the attack, respectively. In addition, we also summarize defense strategies and discuss future directions, offering a broader perspective for researchers in this field. In order to address the security issues in federated recommender systems, we need to continuously explore effective defense mechanisms to ensure the security of the system and provide users with reliable recommendation services.

Reasoning over Knowledge Graphs (KGs) plays a pivotal role in knowledge graph completion or question answering systems, providing richer and more accurate triples and attributes. As numerical attributes become increasingly essential in characterizing entities and relations in KGs, the ability to reason over these attributes has gained significant importance. Existing graph-based methods such as Graph Neural Networks (GNNs) and Knowledge Graph Embeddings (KGEs), primarily focus on aggregating homogeneous local neighbors and implicitly embedding diverse triples. However, these approaches often fail to fully leverage the potential of logical paths within the graph, limiting their effectiveness in exploiting the reasoning process. To address these limitations, we propose ChainsFormer, a novel chain-based framework designed to support numerical reasoning. Chainsformer not only explicitly constructs logical chains but also expands the reasoning depth to multiple hops. Specially, we introduces Relation-Attribute Chains (RA-Chains), a specialized logic chain, to model sequential reasoning patterns. ChainsFormer captures the step-by-step nature of multi-hop reasoning along RA-Chains by employing sequential in-context learning. To mitigate the impact of noisy chains, we propose a hyperbolic affinity scoring mechanism that selects relevant logic chains in a variable-resolution space. Furthermore, ChainsFormer incorporates an attention-based numerical reasoner to identify critical reasoning paths, enhancing both reasoning accuracy and transparency. Experimental results demonstrate that ChainsFormer significantly outperforms state-of-the-art methods, achieving up to a 20.0% improvement in performance. The implementations are available at https://github.com/zhaodazhuang2333/ChainsFormer.

Evolving graphs are prevalent in real-world systems where nodes and edges continuously expand over time, resulting in persistent topological changes across temporal snapshots. As the graph expands, the increasing structural discrepancy between earlier and later timesteps often leads to poor generalization-a phenomenon widely recognized as Temporal out-of-distribution (TOOD) generalization. This issue has spurred growing research interest in generalized graph learning. Existing approaches often rely on modeling latent environments or incorporating structural causal reasoning, but they frequently face scalability limitations when applied to rapidly expanding graphs. In this work, we carry out a detailed analysis of temporal drift in evolving graphs and identify two complementary forms: structural role drift reflected by changes in degree proportions, and semantic drift measured by embedding divergence. To capture this graph evolution, we propose PromptEG, a scalable prompt-based framework that generates node-specific prompts using degree signals and KL-divergence between representations. These prompts are optimized via consistency and contrastive objectives, and injected into GNNs to enhance temporal generalization. Experiments on largescale evolving graphs-containing up to 500 K edges and tens of thousands of nodes-demonstrate that PromptEG achieves strong long-horizon generalization while maintaining high computational efficiency.

With the explosive growth of online multimodal content, multimodal recommender systems(MRSs) have brought significant benefits to multimedia platforms. As MRSs evolve, many studies incorporate advanced technologies like graph neural networks(GNNs) and self-supervised learning(SSL), achieving remarkable results. However, these efforts still suffer from the quality disparity problem. It refers to the mixture of high and low quality across items' multiple modalities, owing to disparities in construction costs or design levels. These low-quality modalities often lack crucial details or introduce noise to the depiction of item, leading to insufficient or polluted item representation. Therefore, we propose a novel framework R2MR: Review and Rewrite Modality for Recommendation to tackle this issue. Specifically, R2MR is composed of two key components: Modality Reviewer and Modality Rewriter. The Modality Reviewer introduces a Consensus Review Mechanism. It performs perspective decomposition based on user representations and learns the consensus quality scores for modalities from diverse perspectives of multiple users. The Modality Rewriter proposes a Latent Mapping Model, which improves the quality of inferior modalities by learning various mapping patterns from high-quality modalities. Comprehensive experiments across three benchmark datasets reveal that R2MR substantially outperforms state-of-the-art methods, achieving an average improvement of 9.20%. The implementations are available at https://github.com/gutang-97/R2MR.

Accurately reconstructing the global ocean deoxygenation over a century is crucial for assessing and protecting marine ecosystem. Existing expert-dominated numerical simulations fail to catch up with the dynamic variation caused by global warming and human activities. Besides, due to the high-cost data collection, the historical observations are severely sparse, leading to big challenge for precise reconstruction. In this work, we propose OxyGenerator, the first deep learning based model, to reconstruct the global ocean deoxygenation from 1920 to 2023. Specifically, to address the heterogeneity across large temporal and spatial scales, we propose zoning-varying graph message-passing to capture the complex oceanographic correlations between missing values and sparse observations. Additionally, to further calibrate the uncertainty, we incorporate inductive bias from dissolved oxygen (DO) variations and chemical effects. Compared with in-situ DO observations, OxyGenerator significantly outperforms CMIP6 numerical simulations, reducing MAPE by 38.77%, demonstrating a promising potential to understand the “breathless ocean” in data-driven manner.

Graph Neural Network (GNN) has demonstrated extraordinary performance in classifying graph properties. However, due to the selection bias of training and testing data (e.g., training on small graphs and testing on large graphs, or training on dense graphs and testing on sparse graphs), distribution deviation is widespread. More importantly, we often observehybrid structure distribution shiftof both scale and density, despite of one-sided biased data partition. The spurious correlations over hybrid distribution deviation degrade the performance of previous GNN methods and show large instability among different datasets. To alleviate this problem, we proposeOOD-GMixupto jointly manipulate the training distribution withcontrollable data augmentationin metric space. Specifically, we first extract the graph rationales to eliminate the spurious correlations due to irrelevant information. Secondly, we generate virtual samples with perturbation on graph rationale representation domain to obtain potential OOD training samples. Finally, we propose OOD calibration to measure the distribution deviation of virtual samples by leveraging Extreme Value Theory, and further actively control the training distribution by emphasizing the impact of virtual OOD samples. Extensive studies on several real-world datasets on graph classification demonstrate the superiority of our proposed method over state-of-the-art baselines.

Model-driven data acquisition is one of the strategies utilized to save sensor node energy in Wireless Sensor Networks (WSNs), which suppresses data transmission by running one synchronized prediction model at both the sensor and sink node, and only when the predicted value deviates far from the real value should the sensor node transmit the sensed data to the sink node. In this paper, we propose a novel online model-driven data acquisition method which runs two prediction models on the sensor node simultaneously. Specifically, one model is updated online using stochastic gradient descent (SGD) learning algorithm once a new sensor data is available and the other is used to predict sensor value and updated with the former one when it is time for model re-training. The collaborative working of these two models, together with the SGD learning algorithm, solve two main problems of existing methods: data transmission during off-line model re-training and high resource requirement for model update. Extensive experiments are performed to verify the benefits of our method over two existing methods based on more than 20000 data records from three data sets. The experiment results demonstrate that up to 96% of data transmission are reduced by our method while remaining user defined data accuracy, which outperforms the compared methods in terms of energy usage and data accuracy.

Dynamic spectrum access (DSA) networks can significantly improve the networks performance and efficiency of spectrum utilization. To harness this capability, spectrum sensing is a fundamental problem for DSA networks. This demonstration shows our wideband compressed spectrum sensing (WCSS) platform. This platform supports the controllable spectrum environment, high performance and low cost wideband spectrum sensing based USRP2, powerful and flexible compressed sensing computing platform based MATLAB. With this platform, we can research the wideband compressed spectrum sensing algorithm by actual and controllable spectrum environment.

Facing the still unclear growth regularity of blooms, we propose rapid lightweight estimating method for the discovery of algae blooms. An application-specific approximate algorithm for timely dispatching salvaging boats is designed to minimize the total cleanup time.

Cardiovascular disease (CVD) has become the primary killer worldwide and is expected to cause more deaths in the future. Prediction and prevention of CVD have therefore become important social problems. Many groups have developed prediction models for asymptomatic CVD by classifying its risk based on established risk factors (e.g., age, sex, etc.). More recently, studies have uncovered that many genetic variants are associated with CVD outcomes/traits. If treated as single or multiple risk factors, the genetic information could improve the performance of prediction models as well as promote the development of individually tailored risk models. In this paper, eligible genome-wide association studies for CVD outcomes/traits will be overviewed. Clinical trials on CVD prediction using genetic information will be summarized from overall aspects. As yet, most of the single or multiple genetic markers, which have been evaluated in the follow-up clinical studies, did not significantly improve discrimination of CVD. However, the potential clinical utility of genetic information has been uncovered initially and is expected for further development.

In order to achieve the high motion precision of the macro-micro parallel manipulator system for a large radio telescope, its real-time motion planning is dealt with in this paper. To uniquely determine the spatial position and orientation of the cable driven parallel manipulator, the real-time optimization is conducted according to the principle of uniform tension in the six driving cables. Synchronized points and the “judge and wait” mechanism ensure the continuity and synchrony of the trajectory tracking of the two parallel manipulators. The preadjustment of the micro parallel manipulator reduces the drastic dynamical coupling as a result of the fast movement of the micro manipulator. Besides, over-workspace avoidance procedure is introduced to restrict the micro parallel manipulator not to exceed it workspace. Experimental results validate the high precision of the macro-micro parallel manipulator system with supervisory controller and the motion prediction algorithm.

Traditional Chinese Pulse Diagnosis is a convenient and noninvasive method for disease diagnosis and healthcare. We have designed and implemented a Chinese wrist-pulse retrieval system based on the principle of Traditional Chinese Pulse Diagnosis (TCPD), called EasiCPRS. It is designed to be small in size, low in cost, with flexibility in deployment, and simplicity in operation. The contributions of this work are: 1. The wrist-pulse at "cun, guan and chi"points over the radial artery are obtained by applying a moderate and adjustable taking pressure during wrist-pulse retrieval. 2. A wrist-pulse signal conditioning circuit and a robust external taking pressure control algorithm are designed to overcome low signal-to-noise ratio (SNR). 3. A lightweight algorithm for wrist-pulse feature extraction is achieved on a resource-constrained platform to economize energy and bandwidth.

Nowadays, the harmful blue-green algae blooms on lakes or streams threaten the daily life of millions of people in China. In this paper, we demonstrate the sensor network system we built on Lake Tai for the surveillance and cleanup of the algae blooms which is at work in Wuxi City, Jiangsu Province. We designed the sensor device and algorithm to monitor the algae bloom and estimate the bloom area. When the bloom area goes beyond the threshold, the salvaging boats are automatically dispatched to the scene to clean up the bloom for recycling utilization according to a mechanism in a scalable coordination fashion. The dispatching mechanism includes to consider the locations of salvaging boats, the number of available salvaging boats and the facility status in eight algae harvesting factories around the lake. The system also balances workloads of factories to achieve an overall high working efficiency. We develop a GIS-based management website for the end user to monitor the running of the whole system. All the sensing stations, the real time performance of the algae harvesting factory, the locations of salvaging boats and the automatically generated cleanup schedule are displayed in the map.

This demonstration shows our versatile heterogeneous sensor networks testbed. This testbed supports high speed embedded WLAN node, medium speed IEEE802.15.4 node, wired computer and so on with powerful supervision platform. The unique features of the testbed are as follows: (1) Supporting heterogeneous sensor networks; (2) Owning powerful testbed nodes; (3) Emulating the interference and noise with the USRP2 replaying the real signal.

This paper presents an overview of our demonstration of a real application of a WSN system for relic protection applied in the Forbidden City of China. We briefly introduce the EasiNet hardware platform we have developed and the corresponding mesh-architecture of the system constructed. We present a sensor deployment optimization tool based on ant colony optimization technology (DT-ACO) to meet the requirements of network connectivity, sensing coverage and a minimized overall cost of the system. We describe a novel power-aware cross-layer scheme (PACS) to facilitate adjustable system lifetime and surveillance accuracy. We also show some detailed results of the application case.

There is a growing need for medical care resources when people pay more and more attention to their health conditions. The hospitals, however, can not satisfy all those ever-increasing requirements. There is hence a contradiction between the needs of the medical care and resource availability. A remote telemedical care system is a possible optimistic way towards solving this problem with the support of emerging new technologies. In this paper, we present the development of a wireless networked Chinese telemedicine system for remote pulse information monitoring, namely WNCTs, based on the principle of Chinese pulse diagnosis in Chinese medicine (CPD). The proposed wearing system may collect the pulse condition, analyze the information by data fusion and provide diagnostic results. The distinctive features of the proposed wireless system include: small in size, light in weight, multiple of type pulse detection and networked users' remote management. In this paper we provide the design and implementation of the wearing wireless networked Chinese telemedicine system. We also demonstrate the use of the system in pulse information retrieval. Our results indicate significant practical feasibility of the networked telemedicine system.

This paper describes a wireless sensor network for mine safety monitoring. Based on the characteristics of underground mine gallery and the requirements for mine safety monitoring, we proposed a distributed heterogeneous hierarchical mine safety monitoring prototype system, namely HHMSM This system is capable of monitoring methane concentration, and locating miner. We proposed a novel overhearing-based adaptive data collecting scheme which exploits the redundancy and correlation of the sampling readings in both time and space to reduce traffic and control overhead with a well-bounded offset error for large-scale sensor networks. This mechanism is easy to implement and low-cost compared to other more theoretically based mechanisms such as Kalman filter. Experimental results show that HHMSM achieves better performance on flexibility, correctness, coverage, and lifetime compared with other existing wireless mine safety monitoring systems.

In March 2004, recognizing the importance of sensor networks, the Chinese Academy of Sciences and the Hong Kong University of Science and Technology launched a joint effort to investigate both fundamental and practical research issues in sensor networks. The goal of this research is to build lightweight optimized sensor systems on a massive scale, namely the BLOSSOMS project. The objective of this research project is to identify research issues at all levels from practical applications down to the design of sensor nodes. This paper reports the status of the project as of April 2005. First, other than making MOTE-compatible sensor nodes, this project has studied the hardware and software co-design and the associated "sensor node on a chip" technology with the aim to make sensor nodes small in size, light in weight, cheap in cost, and low in power consumption. Second, additional toolkits for simulation/emulation and evaluation of sensor networks have been developed to support research at different levels. Third, a number of applications have been investigated and implemented since the launch of this project, and this paper will address two of them. One is an embedded remote health care system based on wireless sensor network technologies to introduce a scalable wireless personal medical network around human's body. The other is a moving object counting system based on an ultrasonic sensor network, which can be used but not limited to crowd estimation and traffic monitoring and coordination.