Explore the vast range of titles available.

After ischemic stroke, survivors experience motor dysfunction and deterioration of memory and cognition. These symptoms are associated with the disruption of normal neuronal function, i.e., the secretion of neurotrophic factors, interhemispheric connections, and synaptic activity, and hence the disruption of the normal neural circuit. Exercise is considered an effective and feasible rehabilitation strategy for improving cognitive and motor recovery following ischemic stroke through the facilitation of neuroplasticity. In this review, our aim was to discuss the mechanisms by which exercise-induced neuroplasticity improves motor function and cognitive ability after ischemic stroke. The associated mechanisms include increases in neurotrophins, improvements in synaptic structure and function, the enhancement of interhemispheric connections, the promotion of neural regeneration, the acceleration of neural function reorganization, and the facilitation of compensation beyond the infarcted tissue. We also discuss some common exercise strategies and a novel exercise therapy, robot-assisted movement, which might be widely applied in the clinic to help stroke patients in the future.

Most acute strokes are ischemic, and subsequent neuroinflammation promotes further damage leading to cell death but also plays a beneficial role by promoting cellular repair. Neutrophils are forerunners to brain lesions after ischemic stroke and exert elaborate functions. While neutrophil extracellular traps (NETs) possess a fundamental antimicrobial function within the innate immune system under physiological circumstances, increasing evidence indicates that NETosis, the release process of NETs, occurs in the pathogenic process of stroke. In this review, we focus on the processes of NET formation and clearance, the temporal and spatial alterations of neutrophils and NETs after ischemic damage, and how NETs are involved in several stroke-related phenomena. Generally, NET formation and release processes depend on the generation of reactive oxygen species (ROS) and the activation of nuclear peptidylarginine deiminase-4 (PAD4). The acid–base environment, oxygen concentration, and iron ions around the infarct may also impact NET formation. DNase 1 has been identified as the primary degrader of NETs in serum, while reactive microglia are expected to inhibit the formation of NETs around ischemic lesions by phagocytosis of neutrophils. The neutrophils and NETs are present in the perivascular space ipsilateral to the infarct arising after ischemic damage, peaking between 1 and 3 days postischemia, but their location in the brain parenchyma remains controversial. After the ischemic injury, NETs are involved in the destruction of neurological function primarily by disrupting the blood–brain barrier and promoting thrombosis. The potential effects of NETs on various ischemic nerve cells need to be further investigated, especially in the chronic ischemic phase.

Differential evolution (DE) is an efficient population-based search algorithm with good robustness, but it faces challenges in dealing with Large-Scale Global Optimization (LSGO). In this paper, we proposed an improved multi-population differential evolution with best-random mutation strategy (called mDE-brM). The population is divided into three sub-populations based on the fitness values, each sub-population uses different mutation strategies and control parameters, individuals share different mutation strategies and control parameters by migrating among sub-populations. A novel mutation strategy is proposed, which uses the best individual and a randomly selected individual to generate base vector. The performance of mDE-brM is evaluated on the CEC 2013 LSGO benchmark suite and compared with 5 state-of-the-art optimization techniques. The results show that, compared with other contestant algorithms, mDE-brM has a competitive performance and better efficiency in LSGO.

Differential evolution (DE) is an efficient population-based search algorithm for solving numerical optimization problems. However, the performance of DE is very sensitive to the choice of mutation strategies and their associated control parameters. In this paper, we propose a self-adaptive multi-population differential evolution algorithm, called SAMDE. The population is randomly divided into three equally sized sub-populations, each with different mutation strategies. At the end of each generation, all sub-populations are updated independently and recombined. Each sub-population uses an adaptive mechanism for selecting how current generation control parameters are generated. An improved mutation strategy, “rand assemble/1”, is proposed, its base vector is composed proportionally of three randomly selected individuals. The performance of SAMDE is evaluated on the suite of CEC 2005 benchmark functions. A comparative study is carried out with other state-of-the-art optimization techniques. The results show that SAMDE has a competitive performance compared to several other efficient DE variants.

In response to the problem of neglecting the periodic and global characteristics of sequence data when predicting PM2.5 concentrations via machine learning models, a PM2.5 concentrations prediction model based on feature space reconstruction and multihead self-attention gated recurrent unit (FSR-MSAGRU) is proposed in this study. First, the raw sequence data are subjected to frequency spectrum analysis to determine the period value of the PM2.5 sequence data. Subsequently, the seasonal trend decomposition procedure based on loess (STL) is employed to capture the periodicity and trend information in the PM2.5 sequence data. Then, the feature space of the PM2.5 sequence data is reconstructed using the raw PM2.5 sequence data, decomposed seasonal components, trend components, and residual components. Finally, the reconstructed feature data are input into multihead self-attention gated recurrent unit (MSAGRU) with the ability to capture global feature information to predict PM2.5 concentrations. Favorable prediction results were attained by the proposed FSR-MSAGRU model across 6 distinct experimental datasets, with a PCC exceeding 0.98 and a decrease in the prediction accuracy metric SMAPE of at least 68% compared to that of the GRU model. Comparative experimental results with 13 reference models demonstrate that the proposed model exhibits better prediction performances and stronger generalization abilities.

Using historical information from traffic accidents to predict accidents has always been an area of active exploration by researchers in the field of transportation. However, predicting only the occurrence of traffic accidents is insufficient for providing comprehensive information to relevant authorities. Therefore, further classification of predicted traffic accidents is necessary to better identify and prevent potential hazards and the escalation of accidents. Due to the significant disparity in the occurrence rates of different severity levels of traffic accidents, data imbalance becomes a critical issue. To address the challenge of predicting extremely imbalanced traffic accident events, this paper introduces a predictive framework named ReMAHA–CatBoost. To evaluate the effectiveness of ReMAHA–CatBoost, we conducted experiments on the US–Accidents traffic accident dataset, where the class label imbalance reaches up to 91.40 times. The experimental results demonstrate that the proposed model in this paper exhibits exceptional predictive performance in the domain of imbalanced traffic accident prediction.

In machine learning (ML) algorithms, neural networks (NNs) can effectively learn the mapping between initial condition errors and system states through training. To reduce model errors in data assimilation, this work proposes an optimization strategy for the data assimilation (DA) process based on ML methods. A novel hybrid deep learning approach is introduced, which combines bidirectional long short‐term memory (BiLSTM) networks and gated recurrent units (GRUs) in a serial processing manner. First, the BiLSTM‐GRU hybrid method is trained by modeling the residuals between the system's observational data and the ensemble Kalman filter (EnKF) assimilation results, which are used as inputs to learn the error correction process. Second, the Huber loss function is used to quantify the error, and the network model's parameters are updated on the basis of the loss feedback to correct model errors in the data assimilation system. Additionally, the hybrid model incorporates nudging by introducing a certain amount of “relaxation” in each iteration, gradually approaching the actual solution. Ultimately, comprehensive experiments demonstrate that the BiLSTM‐GRU hybrid, even after strengthening traditional EnKF with inflation, localization, and stabilization, consistently achieves superior robustness across varying dynamical regimes, ensemble sizes, observation densities, and noise levels, highlighting its scalability for nonlinear chaotic systems. Plain Language Summary Machine learning has great potential for research on data assimilation methods. This study improves data assimilation by using a hybrid machine learning model combining BiLSTM and GRU networks. The model learns to correct errors by analyzing differences between observations and predictions from the EnKF via a robust error method (Huber loss) and gradual refinement (“nudging”). Tests on noisy and sparse data from the Lorenz‐96 system show that the method outperforms the traditional EnKF and handles considerable noise and sparse observations with greater robustness and adaptability. Key Points The BiLSTM‐GRU hybrid model combines bidirectional dynamics and gating mechanisms, outperforming single models and EnKFs in complex data assimilation Replacing MSE with Huber loss makes the model more robust to outliers and noise, improving stability and accuracy ReLU and dense layers boost deep feature extraction, improving adaptability and precision in time—series modeling compared to EnKFs and RNNs

Quantitative evaluation and training of the reach-to-target ability in stroke patients are needed for postdischarge rehabilitation, which can be achieved using a motion-tracking system. However, most of these systems are either costly, involve sophisticated parameter interpretation, or are not designed for rehabilitation. We developed an interactive reach-to-target assessment and training system (IRTATS) based on a camera and three marker straps to detect tracking signals. IRTATS supports audiovisual feedback, personal goal setting, and use in a small clinic or home without the internet. This study aims to evaluate the reliability, validity of IRTATS, and its measurement accuracy of the range of motion (ROM). Ninety-nine stroke patients and 20 healthy adults were recruited for the study. Kinematic variables and active joint ROM (AROM) were assessed using IRTATS. The AROM was measured by a universal goniometer, and scores from multiple clinical scales concerning motor and activity capability were calculated. Although the AROMs measured by IRTATS and the goniometer did not agree, IRTATS has clinically acceptable reliability and validity. Three variables in IRTATS could discriminate the motor performance of patients and healthy subjects. IRTATS may provide a new supplement to conventional physiotherapy in the assessment of the reach-to-target ability in stroke patients.

Accurate state estimation in complex dynamical systems is crucial for understanding and predicting nonlinear physical processes. However, conventional ensemble-based Kalman filtering methods, which rely on L 2-norm updates, are highly sensitive to outliers and gross observational errors, resulting in degraded performance in chaotic regimes. To address this challenge, we develop the EnTLHF-R, an enhanced Ensemble Time-localized H-infinity Filter that integrates observation error estimation, dynamic quality control, and the Huber loss function. By combining the efficiency of L 2 updates with the robustness of L 1, the proposed framework effectively suppresses the impact of additive and innovation outliers while preserving sensitivity to reliable observations. Simulation experiments using the Lorenz-96 model demonstrate that EnTLHF-R achieves markedly improved stability and accuracy in high-dimensional chaotic systems. These results highlight the method’s potential for advancing robust data assimilation in nonlinear physical systems and improving predictability under adverse observational conditions.

In the central nervous system, microglia are considered as the main cells responsible for phagocytosis, contributing to neural circuit refinement and homeostasis through synapse elimination during development. Microglial phagocytosis is also a crucial process in maintaining adult brain homeostasis and clearing potential toxic factors, which are recognized to be associated with neurodegenerative and neuroinflammatory disorders. Microglia can engulf cellular debris and harmful molecules such as amyloid-β plaques through the expression of specific receptors at their surface and activation of downstream signaling pathways. Furthermore, Physical exercise is an effective therapy for neurorehabilitation which has been shown to be closely related to neuroplasticity and microglia functions including proliferation, activation and phagocytosis. This review focuses on the central regulatory mechanisms related to microglia phagocytosis and the potential role of exercise training in this process.