- Research Article
- 10.1016/j.inffus.2026.104287
Federated learning for spatio-temporal data mining: a survey
- Sep 01, 2026
- Information Fusion
- Wei Huang + 8 more +8
Publications from 2021 to 2026
Showing 10 of 528 papers
Federated learning for spatio-temporal data mining: a survey
A reinforcement learning-based approach for offline and online multi-line bus scheduling problem
Research on temperature estimation and suppression methods for permanent magnets in interior permanent magnet synchronous motor
TR-DQ: Time-Rotation Diffusion Quantization
Diffusion models have been widely adopted in image and video generation. However, their complex network architecture leads to high inference overhead for its generation process. Existing diffusion quantization methods primarily focus on the quantization of the model structure while ignoring the impact of time-steps variation during sampling. At the same time, most current approaches fail to account for significant activations that cannot be eliminated, resulting in substantial performance degradation after quantization. To address these issues, we propose Time-Rotation Diffusion Quantization (TR-DQ), a novel quantization method incorporating time-step and rotation-based optimization. TR-DQ first divides the sampling process based on time-steps and applies a rotation matrix to smooth activations and weights dynamically. For different time-steps, a dedicated hyperparameter is introduced for adaptive timing modeling, which enables dynamic quantization across different time steps. Additionally, we also explore the compression potential of Classifier-Free Guidance (CFG-wise) to establish a foundation for subsequent work. TR-DQ achieves state-of-the-art (SOTA) performance on image generation and video generation tasks and a 1.38-1.89× speedup and 1.97-2.58× memory reduction in inference compared to existing quantization methods.
Read moreA novel superficial temporal artery patency concept of cerebral revascularization for patients with moyamoya disease: a multicenter study.
To summarize the clinical and radiological outcomes of a novel cerebral revascularization technique based on the superficial temporal artery patency concept (STAPC) in patients with moyamoya disease (MMD). A retrospective review was conducted of adult patientswith MMD treated at Beijing Hospital and Nanjing Drum Tower Hospital between January 2019 and December 2021. The cohort comprised 170 patients who underwent superficial temporal artery-middle cerebral artery bypass with encephalo-duro-arterio-synangiosis (EDAS) (STA-MCA/EDAS), and 133 who underwent EDAS alone. Radiological follow-up included computed tomography (CT) angiography (CTA) to assess bypass patency and CT perfusion (CTP) for hemodynamic staging at 3 and 12 months post-revascularization. Clinical follow-up recorded perioperative complications and recurrent stroke eventsthat occurred > 12months postoperativerly. Of the 303 patients, 37 cases (12.21%) had perioperative complications including 27 cases (15.9%) inthe STA-MCA/EDAS group and 10 cases (7.5%) in the EDAS group. Perfusion improvement was observed in 23.76% ofpatients (25.7% in the STA-MCA/EDAS group and 17.8% in the EDAS group) at 3monthspostsurgical, and in40.17% ofpatients (48.68% in the STA-MCA/EDAS group and 24.39% in the EDAS group) at 12monthspostsurgical. Bypass patency was observed in 95.29% ofpatients (96.1% in the STA-MCA/EDAS group and 93.5% in the EDAS group) at 3monthspostoperative, andin 96.43% patients (95.2% in the STA-MCA/EDAS group and 97.1% in the EDAS group) at 12monthspostoperative. Of the 249 patients with a median follow-up period of 50months (range 12-70months), 40 cases (16.06%, 3.85% per year) had recurrent stroke events including 22 (15.41%, 3.69% per year) in the STA-MCA/EDAS group and 18 (16.67%, 3.91% per year) in the EDAS group. Cerebral revascularization using STAPC is an acceptable surgical strategy for preventing stroke recurrence in patientswith MMD.
Read moreAdvancing Fire Management in the New Era: Toward Comprehensive Refinement and Inclusive Mobilization for Enhanced Operational Efficiency
This study examines the perceptions and expectations of diverse stakeholders in society with respect to contemporary fire management. By tracing the evolutionary trajectory of fire management practices in recent years, the paper proposes a progressive model that transitions from experience-based approaches to evaluation-driven upgrading, and ultimately toward comprehensive refinement, decomposition, and systematic implementation. In parallel, the study synthesizes the roles, responsibilities, and obligations undertaken by relevant parties during the design and construction phases [1], alongside practical insights derived from fire management during the operational phase. It is posited that the advancement of fire safety management across various organizational units necessarily depends on the adoption of a comprehensively refined and decomposed management model. This paradigm not only enhances operational precision but also serves as a foundational pathway for institutionalizing fire safety performance.
Read moreInsights into phase evolution, foaming behavior and heavy metal immobilization in the oil-based cutting pyrolysis residues based lightweight building ceramsite production: The role of SiC
Unveiling the impact of ultrasonic frequency pulse current assisted-MIG welding on the microstructure, mechanical properties, and corrosion behavior of 316L stainless steel
Research progress on BCC solid solution type high entropy hydrogen storage alloys
Microstructure refinement in near-α/Ti2AlNb gradient transition zone by laser deposition-hot deformation hybrid manufacturing