Hypergraphs generalize traditional graphs by allowing hyperedges to connect any number of vertices, enabling the representation of complex relationships. Subhypergraph matching queries aim to identify all subhypergraphs within a hypergraph that are isomorphic to a given query hypergraph. However, existing subhypergraph matching methods often fail to leverage the higher-order structure of hypergraphs efficiently. At the same time, these methods typically require traversing each hyperedge in the data hypergraph to find candidate sets for query hyperedges, leading to high computational and time costs. In this paper, we introduce a lossless transformation structure called the hypergraph mapping graph. Leveraging this structure, we propose a hyperedge embedding method that ensures if a hyperedge in the query hypergraph matches one in the data hypergraph, the embedding of the data hyperedge dominates that of the query hyperedge. We then apply spatial indexing to the hyperedge embeddings of the data hypergraph, enabling direct retrieval of query candidate sets without requiring traversal of the data hypergraph. Finally, we present a parallel matching method that capitalizes on GPU parallelism to accelerate the enumeration process. Experimental results demonstrate that our proposed method significantly outperforms existing approaches.

The rapid advancement of Internet of Things (IoT) and big data technologies has increased the demand for storage systems that can sustain high write throughput while efficiently supporting secondary attribute queries. We present GBSI, a GPU-based secondary indexing mechanism for Log-Structured Merge Tree (LSM-tree)-based key-value stores under writeintensive workloads with append-dominated data ingestion. GBSI adopts a decoupled CPU–GPU architecture, enabling independent maintenance of the primary LSM-tree and the secondary index, thereby accelerating secondary attribute queries while preserving primary-index write efficiency. GBSI designs GPUresident index structures to manage the growth of pkey lists, supporting scalable secondary index construction and query processing under high-rate writes. To address the limited memory capacity of a single GPU, GBSI is extended to a distributed design that partitions the secondary index across multiple GPUs using a hash-based routing scheme. We evaluate GBSI using synthetic workloads and real-world industrial datasets, and compare it with representative LSM-tree-based systems, including RocksDB and the decoupled secondary indexing system SineKV. Experimental results show that GBSI improves write throughput by ${1. 0 5} \times$ to ${1. 2 1} \times$ compared with RocksDB and by $\mathbf{1. 0 6} \times$ to 3.36 compared with SineKV. For secondary attribute queries, GBSI delivers up to two orders of magnitude speedup at the 109 data scale, demonstrating its effectiveness for the targeted write-intensive workloads.

Video Database Management Systems (VDMS) face a critical challenge in optimizing query plans: accurate selectivity estimation, which predicts the proportion of data satisfying a predicate. Existing methods rely on fixed sampling rates (e.g., 10

Graphs are widely used to represent vertices and their relationships, with the subgraph counting problem being a crucial challenge in network analysis. Originating from subgraph matching, this problem aims to efficiently determine the number of subgraphs in a data graph that are isomorphic to a given query graph. Since this task is NP-hard, existing approaches are typically divided into exact and approximate methods. Exact methods compute the precise number of isomorphic subgraphs but incur prohibitive computational costs on large graphs. Approximate methods, often based on Graph Neural Networks (GNNs), estimate the subgraph count more efficiently but tend to suffer when there is a large size discrepancy between the data graph and the query graph, leading to inaccurate vertex matching. To address these limitations, we propose a domination-based subgraph counting method (DSCM), which improves vertex matching by using a vertex dominance embedding method. This approach, combined with an attention mechanism, accurately computes subgraph counts. Experimental results show that DSCM outperforms existing methods, providing an efficient solution for subgraph counting in large graphs.

The function-calling agent is obtained by performing agent tuning to the large language model (LLM) on function-calling dataset. However, even state-of-the-art datasets (e.g., xlam-function-calling-60k datasets) still contain numerous misleading examples of low-quality data, wasting significant computational resources and result in an unnecessary carbon footprint. Furthermore, such inductive bad data negatively impacts the performance of the agent. In this paper, we propose a set of scoring criteria specifically tailored to evaluate function-calling data and use these criteria to develop a data filtering framework. By applying this framework to filter out low-quality data, we fine-tuned SuperFC, which demonstrates substantial improvements in both sustainability and performance. The SuperFC-7B training process reduced training time from 455 minutes to 85 minutes, resulting in a 80.02% reduction in carbon footprint. Simultaneously, fine-tuning on high-quality data subsets led to performance improvements of up to 3.68%. Additionally, we provide an in-depth analysis of the causes behind the low quality of synthetic function-calling data, offering valuable insights for future data synthesis in this domain. We have also released a high-quality function-calling dataset, available at: https://github.com/Zire-Young/SuperFC

JointCloud computing emerages as a novel paradim to alleviate the platform lock-in issues faced by traditional cloud computing, but it remains constrained by the autonomy of cloud service providers and their stringent data privacy requirements. Thus, applying federated learning in a JointCloud environment enables secure collaborative computing across geographically distributed devices. However, in practical deployments, federated learning relying on a single global model faces multiple heterogeneity challenges, such as label distribution skew and computational capability heterogeneity. Addressing these challenges, our proposed solution, FedSelf, implements self-distillation within a federated learning architecture. This framework clusters devices according to their computational and storage capabilities, assigning high-resource clients receive larger model training tasks, whereas low-resource ones handle smaller model training. During training of large models, self-distillation technique is applied, thereby enabling collaborative training across models of different sizes. Furthermore, before each training round, clients can dynamically adjust their model sizes based on realtime resource availability, optimizing computational resource allocation and enhancing system stability. Experimental results across multiple datasets demonstrate that FedSelf effectively enables the aggregation of heterogeneous models, achieving an approximate 2% improvement in accuracy compared to baseline methods.

With advancements in positioning technology , the widespread adoption of wireless sensors, numerous wireless handheld and vehicular devices now come equipped with positioning capabilities. This has enabled a variety of new applications and generated large volumes of moving object data. The continuously changing location information of these moving objects requires efficient management in databases. Traditional database systems, which typically assume static attribute values until explicitly updated, face challenges in managing such dynamic, constantly changing location data efficiently. Current moving object indexing structures fall mainly into two categories: grid-based and tree-based indexing. However, each approach has inherent limitations. In this paper, we propose a novel indexing method that combines grid and quadtree structures and utilizes a deep learning model to intelligently determine when leaf nodes should be split or merged. Our method is not just a theoretical concept, but a practical solution that can be applied to a wide range of scenarios. Experimental results demonstrate that our method achieves higher throughput and reduces response times, particularly in skewed moving object distributions, offering significant improvements over existing indexing techniques.

GPUs are being applied widely to batch workloads that benefit from the parallel processing capabilities of GPUs. To enable the processing of concurrent batch-based transactions on GPUs, existing systems build dependency graphs during a pre-execution phase to manage read and write operations. However, as dependency-graph maintenance introduces a sub-stantial overhead, there is a need for more efficient transaction support to exploit the power of GPUs more fully for transaction processing. This paper proposes LTPG, a novel GPU-enabled database system that offers increased versatility and efficiency by eliminating the need for predefined read/write-sets. LTPG employs deterministic optimistic concurrency control to ensure correct transaction execution, thus avoiding the maintenance of dependency graphs. The proposed concurrency control simpli-fies transaction processing workflows and avoids the overhead associated with managing dependency graphs, thus resulting in improved efficiency. LTPG divides a workflow into three stages: execution, conflict detection, and write-back, leveraging the parallelism of GPUs. Moreover, several additional optimization strategies are adopted to improve system performance. Experiments with real-world workloads from two benchmarks verify LTPG can achieve effective improvement in the throughput and latency compared to the leading baselines.

To realize the great flexibility and cost savings for providing location-based service, data owners are incentivized to migrate their data to cloud servers. However, direct data outsourcing to untrusted servers may pose significant privacy risks. This paper proposes a practical privacy-preserving nearest neighbor searching method over encrypted spatial data. We simultaneously protect data and location privacy (access and pattern privacy) by encrypting data using asymmetric scalar-product-preserving encryption (ASPE) and performing computational private information retrieval (CPIR) on encrypted subspace datasets. To mitigate the performance degradation introduced by the combination of ASPE and CPIR, we propose a hierarchical index that enables users to safely obtain encrypted subspace datasets with configurable privacy, where different degrees of privacy can be traded off against query processing performance. Experiments demonstrate that our method outperforms the state-of-the-art method in efficiency while allowing for a flexible trade-off between performance and privacy.

Federated learning, which allows multiple client devices in a network to jointly train a machine learning model without direct exposure of clients' data, is an emerging distributed learning technique due to its nature of privacy preservation. However, it has been found that models trained with federated learning usually have worse performance than their counterparts trained in the standard centralized learning mode, especially when the training data is imbalanced. In the context of federated learning, data imbalance may occur either locally one one client device, or globally across many devices. The complexity of different types of data imbalance has posed challenges to the development of federated learning technique, especially considering the need of relieving data imbalance issue and preserving data privacy at the same time. Therefore, in the literature, many attempts have been made to handle class imbalance in federated learning. In this paper, we present a detailed review of recent advancements along this line. We first introduce various types of class imbalance in federated learning, after which we review existing methods for estimating the extent of class imbalance without the need of knowing the actual data to preserve data privacy. After that, we discuss existing methods for handling class imbalance in FL, where the advantages and disadvantages of the these approaches are discussed. We also summarize common evaluation metrics for class imbalanced tasks, and point out potential future directions.

In order to release the impact of redundant domination testing on querying efficiency of Skyline processing, and to improve the efficiency of Skyline query processing, we expand the domination testing target from points to grids. We also design a cell division-based Skyline query processing method. In this method, we incrementally shrink the cells that contain Skyline points where we can prove that the domination relationship between cells are the same as that between points. By this means, we avoid tremendous redundant domination testing between points and hence improve the efficiency. In our method, cells in the same layer can be processed in parallel. Moreover, the precision can be reached at any level. Therefore, our method is more suitable than existing method in cases that the processing ability is limited, and the precision requirement is not high. The experimental results show that, compared with existing methods, our method can handle more data and response more rapidly. The advantage is more obvious in when high-end GPU or multi-core CPU is available.

通过风能科学与工程课程传统教学和PBL教学模式进行对比,讨论了PBL教学模式及主动学习在课程教学中的效果.采用线下线上结合形式进行课程教学互动和考核.PBL及主动学习模式促进了学生自主学习意识,培养了社会性能力,锻炼了高层次思维.

Skyline查询是一种基本的数据库操作,在推荐系统等多标准决策场景中有许多应用.随着数据量的迅速增长和信息技术的不断发展,原有的高纬数据Skyline查询算法存在着成本高、速度慢的问题.本文提出了新的基于网格划分的Skyline查询算法,该算法有效利用了新型CPU结构多线程能力增强的特点,提高了对硬件的利用效率.在基于网格划分的思想下,数据空间被划分为多个网格,以网格为单位检查支配关系,这样可以按网格进行修剪,而无需检查每个网格中的点.实验结果表明,本文提出的算法在多个方面都优于最先进的Skyline算法,尤其在计算时间上的优势最为明显.

基于位置的服务(location based services,LBS)为人们日常生活带来极大便利的同时也严重地威胁到用户的隐私.为了在保护用户位置隐私的同时降低用户的查询代价,提出了一种基于安全索引的位置隐私保护方法(location privacy protection method based on secure index,LPPMSI).LPPMSI不需要用户搜集数据信息,数据拥有者为虚假位置信息建立安全索引.用户能够安全、快速地检索到虚假位置信息,服务器再执行基于虚假位置的LBS.并且还提出了基于滑动窗口的候选位置选择算法,有效减少候选位置选择时间.实验结果表明,LPPMSI在不改变隐私保护能力的前提下,不需要用户花费大量时间搜集数据分布及查询概率信息.

Location-based service has become one of the essential parts of our daily lives, which raises a problem: how to preserve location privacy. Computational Private Information Retrieval (CPIR) protocol, which does not need a trusted third party, allows users to retrieve data from a service provider without revealing users' query information. However, the CPIR protocol scans the entire data space, which causes high costs and limits the scale of CPIR-based applications. To address this problem, we propose a performance tunable Computational Private Information Retrieval (PTCPIR) model. We divide data space into subspaces and scan a subset of subspaces to achieve better performance while sacrificing some privacy. By tuning the percentage of scanned subspaces, we can achieve a tradeoff between performance and privacy. We design a hierarchical encrypted secure index to support data users to retrieve subspace information safely and efficiently, which facilitates privacy-preserving subspace information retrieval. Furthermore, we propose a novel subspace information retrieval method where the confused query keyword sequence makes servers unable to count query probability. Hence access pattern privacy is protected. Experiments on Sequoia and synthetic datasets demonstrate that our PTCPIR model provides a tunable tradeoff between performance and privacy with high efficiency and good scalability. (C) 2021 Elsevier Inc. All rights reserved.

The pivoted subgraph isomorphism problem is a special subgraph isomorphism problem that focuses on the pivoted nodes rather than the entire subgraphs. The key challenge in adapting existing techniques to the pivoted problem is eliminating their redundant intermediate results. In this paper, we propose a GPU-based pivoted subgraph isomorphism filtering technique, where information of each node is encoded into a series of codes. When performing a pivot subgraph search, the candidate nodes satisfying the coding requirements are collected parallelly on GPU while others are filtered away. Then the final result can be effectively retrieved by a verification process on the filtered nodes. As demonstrated by the experimental results, our method dramatically reduces the processing time of the pivoted subgraph isomorphism problem. Compared to the state-of-the-art GPU-friendly subgraph matching method GpSM which also focuses on filtering effect, the algorithm’s execution time is halved, confirming that our approach can effectively process pivoted subgraph isomorphism queries.

子图同构问题是非确定多项式(NP)完全问题,而轴心子图同构是一种特殊的子图同构问题.针对现在已经有许多高效的子图同构算法,然而对于轴心子图同构问题目前并没有基于GPU的搜索算法,且通过改造已有的子图同构算法来解决轴心子图匹配问题会产生大量不必要的中间结果这一问题,提出了一种基于GPU的轴心子图同构算法.首先,通过一种新颖的多编码树方式,利用节点的标签、度以及节点邻居的结构特征的组合对节点进行编码,并在GPU上对查询图节点并行地进行剪枝,从而明显地减小数据图候选节点所生成的搜索空间树的尺寸;然后,逐层访问查询图节点的候选节点,过滤掉不满足的节点;最后,验证得到的子图是否是查询图的同构子图,从而高效地完成轴心子图同构搜索.实验结果表明,与GPU友好子图匹配(GpSM)算法相比,所提算法的执行时间降低了二分之一,且该算法能够高效地执行轴心子图同构搜索并且具有可扩展性.所提轴心子图同构算法可以减少解决轴心子图同构问题所需的时间,同时降低了GPU内存消耗,提升了算法的性能.

Skyline computation is an essential database operation that has many applications in multi-criteria decision making scenarios such as recommender systems. Existing algorithms have focused on checking point domination, which lack efficiency over large datasets. We propose a grid-based structure that enables grid cell domination checks. We show that only a small constant number of cells need to be checked which is independent from the number of data points. Our structure also enables parallel processing. We thus obtain a highly efficient parallel skyline algorithm named SkyCell, taking advantage of the parallelization power of graphics processing units. Experimental results confirm the effectiveness and efficiency of SkyCell -- it outperforms state-of-the-art algorithms consistently and by up to over two orders of magnitude in the computation time.

The Graph Neural Network (GNN) is a rising graph analysis model family that encodes node features into low-dimensional representation vectors by aggregating local neighbor information. Nevertheless, the performance of GNNs is limited since GNNs are trained only over predictions of the labeled data. Hence, effectively incorporating a great number of unlabeled nodes into GNNs will upgrade the performance of GNNs. To address this issue, we propose a Contrastive Semi-supervised learning based GNN (CSGNN) that improves the GNN from extra supervision predicted by contrastive learning. Firstly, CSGNN utilizes multi-loss contrast to learn node representations via maximizing the agreement between nodes, edges and labels of different views. Then, a semi-supervised fine-tuner learns from few labeled examples while making the best use of unlabeled nodes. Finally, we introduce the knowledge distillation based on label reliability, which further distills the node labels predicted by contrastive learning into the GNN. Experimentally, CSGNN effectively improves the classification performance of GNNs and outperforms other state-of-the-art methods in accuracy over a variety of real-world datasets.

对不同年级学生同时进行了冶金概论课程PBL教学改革实践.讨论了学生已有知识经验和教学背景对课程PBL教学中项目完成效果的影响,为不同年级全面开展PBL教学提供支持和经验积累.实践表明,已有知识经验和教学背景对项目的选题影响最大,但在学生能力和高层思维培养上没有差别,PBL的教学模式可以适应各年级知识体系,同时在不同年级开展教学活动.