Concurrency and Computation-Practice & Experience最新文献

The prevalence of pulpal and periapical diseases exceeds 50% in the general population. Root canal treatment is currently recognized as the gold standard treatment, in which precise measurement of working length (WL) is critical for treatment success. In this study, a total of 100 eligible extracted teeth were collected and scanned using cone-beam computed tomography (CBCT) to obtain high-resolution three-dimensional images. For WL calculation, we employed a V-net-based segmentation network for simulated paths of the root canal file, incorporating an encoder–decoder structure and a multi-task learning strategy, and achieved a sensitivity of 94.7%. Ablation studies revealed that integrating the decoder's mask branch, key point branch, and boundary branch significantly improved the segmentation accuracy. The WL calculation comprised three stages: skeleton extraction and noise suppression, branch extraction and generation of the simulated paths of the root canal file, and length calculation based on three-dimensional spline curves. The model achieved an average prediction error of 0.28 mm and an accuracy of 86.67% in WL prediction. These findings indicate that this V-net-based multi-branch framework for precise WL estimation from CBCT holds substantial clinical application potential. Future work will focus on enhancing generalization and addressing challenges posed by calcified or anatomically complex root canals.

A competency-based approach supported by personalized learning paths and prompt feedback accelerates skill development by continuously adapting to learners' needs and maintaining high levels of engagement. Capturing and understanding learner competency development through interaction data offers the potential for early intervention and optimized educational design, yet introduces challenges related to scalability and complexity. We present TracePath, a novel graph-based framework that models learner trajectories as directed graphs, where nodes correspond to competencies or learner states and edges denote transitions such as validation or rejection events. This approach uncovers common learning pathways, identifies bottlenecks, and supports predictive analytics. At the core, a generic metamodel formalizes Competency Transition Graphs (CTGs), enabling comprehensive graph-based analytics implemented over a hybrid polystore architecture that integrates both relational and NoSQL databases. Our design decouples data extraction from graph exploration, allowing efficient querying, clustering, and pattern matching to deliver timely and explainable learning insights. Empirical validation using real-world data from the écri+ e-certification project demonstrates TracePath's effectiveness in providing scalable, dynamic, and low-latency learning analytics to support personalized education.

The LSM-tree based on extendible hashing has been adopted in key-value storage systems due to its high write throughput, good scalability, and balanced read-write performance. However, it faces several challenges when accessing SSTables, including low cache efficiency, uneven data distribution across hash buckets, and frequent directory expansions. To address these issues, this paper proposes a cache-friendly LSM-tree based on extendible hashing. To solve the problem that similar keys in SSTables are not stored adjacently in SSTables, a key-to-value mapping strategy based on Locality-Sensitive Hashing (LSH) is employed. Second, to address the uneven data distribution across hash buckets in extendible hashing, a logically uniform extendible hashing scheme is designed, along with a novel HTable structure to replace the traditional SSTables in LSM-tree. In addition, an HTable indexing strategy based on the LSH-Cuckoo filter is proposed to accurately locate the target HTable. Based on the Intel Optane DC Persistent Memory driver, a prototype of a cache-friendly key-value storage system named DLMS was implemented on non-volatile memory (NVM), and evaluated using the YCSB benchmark. Experimental results show that, compared to the LSM-tree-based storage system RocksDB, DLMS achieves an average improvement of 9.8% in read throughput and 11.9% in write throughput, while reducing insertion latency by 7.8%.

The existence of Bushfire is a significant problem. So not only are environmental threats to life and society a risk, but also the threatening of the safety of all. Thus it is extremely critical to detect these fires in time. This identification can assist in the successful application of fire management. Existing approaches might not be effective under conditions such as bad weather. A decrease in visible light also can affect accuracy. These techniques focus on the identification of fires during early stages. We introduce an innovative method in this study. The early detection of bushfires is achieved through UAVs. Each drone in the network has a thermal image camera onboard to provide real-time surveillance of the assigned area. The drones are also equipped with sensors that detect smoke and other indications of potential fire. A control network receives up-to-the-minute information from drones. Machine learning algorithms are used to interpret this data. The aim is to detect potential fires. The above detailed evaluations were captured using the eBee X and DJI Matrice 300 RTK. This study was carried out in a dedicated area provided by a National Park and involved the use of fixed and multirotor UAVs. State of the art sensor system could be successfully deployed on board the UAVs. The integrated technology was comprised of a Sequoia+ multispectral camera, H20T Quad Sensor and PGS 813 Gas Sensor. These were test burns, including controlled prescribed burning and fire simulations. They highlighted the system's impressive ability to detect bushfires in their early stages. The system has provided a quantum leap in preventive fire management. And it has heightened cost-mindedness in real-world settings. This enhancement is due to its high-quality monitoring and precise detection of fire signs. The proposed method might be a useful and powerful tool for proactive fire fighting and early warning, such as fighting wildfire and related scenarios.

Accurate discrimination of edible mushrooms is critical for food safety, yet it remains challenging due to the high visual similarity between edible and toxic species, leading to frequent misidentification by conventional vision-based systems. In this paper, we propose a novel Belief Measure-Weighted Fusion enhanced model that integrates Dempster–Shafer evidence theory into mushroom recognition for the first time. Our enhanced model consists of two core components: a Probabilistic Classification Module with Parallel Color Representation that extracts multidimensional features through complementary color spaces and produces diversified soft predictions; and a multisource classification decision fusion (MCDF) Module that effectively reconciles and integrates conflicting evidence from these predictions. A key innovation within MCDF is the belief cosine similarity coefficient (BCSC), which quantitatively assesses inter-evidence conflict, enabling an adaptive evidence fusion method that enhances robustness and decision reliability. Extensive experiments on a hybrid dataset containing challenging species show consistent performance gains across six mainstream deep networks, demonstrating strong generalization. This work not only bridges evidence theory with practical mushroom identification but also offers a transferable framework for recognizing visually similar species in wild food contexts.

We introduce $��{�\text{DR}�}_{�\pi �}^{�\phi �}�$, a novel routing calculus specifically designed to dynamically update routing tables using distance-vector routing updates in distributed computing networks. This calculus features a three-tier syntactic structure where routers form an undirected graph representing their connectivity, which does not need to be a complete graph. The nodes hosting processes are directly connected to certain routers. A key aspect of our calculus is the periodic and real-time exchange of updates of the routing table between adjacent routers, ensuring that routers consistently provide the optimal path for message transmission between communicating processes. Our calculus offers a more accurate representation of actual distributed networks within a process-algebraic framework. To validate our approach, we demonstrate that the equivalence between well-formed configurations in reduction semantics can be achieved through bi-simulation-based equivalence over labeled transition systems (LTSs), and vice versa.

Multi-Object Tracking (MOT) is promising for UAV applications but faces challenges such as small targets, occlusions, motion blur, and cluttered backgrounds. This work proposes a one-shot MOT framework with a retrospective matching architecture for UAV ground tracking. Using FairMOT as the baseline avoids two-stage redundancy, enhancing UAV deployment efficiency. The lightweight retrospective matching network, comprising feature extraction, sparse graph tracking, and a refinement module, leverages historical information to recover tracking failures, improving continuity and accuracy with low computational cost. Experiments on public benchmarks show superior tracking performance and real-time efficiency.

This study presents DFSplat, a feed-forward 3D Gaussian Splatting model utilizing depth feature fusion for the high-quality reconstruction of a 3D scene from sparse multiview data and the production of novel view images. DFSplat enhances the robustness of depth prediction and the quality of geometric reconstruction by incorporating a pre-trained monocular depth estimation module into the multiview feature matching branch, thereby addressing the limitations of current methods for multiview depth estimation in complex scenes. The method employs a content-guided attention (CGA) module to adaptively integrate monocular depth features with multiview cost-volume features, addressing the fusion difficulty arising from the disparity in encoding between low-level and high-level features. Experiments conducted on the extensive RealEstate10K and ACID data sets demonstrate that DFSplat surpasses current methodologies in PSNR, SSIM, and LPIPS measures, attaining state-of-the-art performance. The innovation integrates the global consistency of monocular depth estimation with the local precision of multiview matching, optimizing 3D Gaussian parameters prediction through an efficient fusion strategy, thereby offering a novel approach for high-quality scene reconstruction in sparse view scenarios.

Accurate forecasting of air pollutant concentrations is crucial for environmental protection and public health. This paper proposes STAG, a novel spatiotemporal architecture for multi-step air quality prediction. The model features a symmetric inner–outer layer structure that integrates Stacked Fluctuation-Weighted Multi-Head Self-Attention (SF-MHSA)and Graph Networks (GNs), effectively capturing complex temporal patterns and inter-pollutant relationships while maintaining computational efficiency through parameter sharing. We introduce a multi-order loss function incorporating first and second-order difference errors to enhance prediction stability and trend accuracy. Additionally, a stationarity correction mechanism is designed to mitigate distribution drift in long-term forecasting. Comprehensive interpretability analyses validate the model's capability to learn meaningful spatiotemporal dependencies. Forecasting results on Shenzhen air quality data further demonstrate that STAG achieves superior performance against state-of-the-art benchmarks in both accuracy and efficiency (lowest $��A�Q�I�$ prediction MSE 366.68 vs. the second best of 477.76). The proposed framework provides an effective solution for air quality forecasting with potential applications in other environmental monitoring domains.