Explore the vast range of titles available.

Correlation filters (CFs) manifest the advantage of high computational efficiency in target tracking, especially in unmanned aerial vehicles (UAVs) target tracking. CF-based trackers update their models whenever necessary to improve the discrimination of the models. However, most existing update strategies only use the single-frame information to evaluate the reliability of the current tracking result, while neglecting the temporal relationship between frames, which could lead to the model degradation problem. In this paper, an Efficient spatial-temporal UAV visual tracker with a Temporal Enhancement model Update strategy (ETEU) is proposed. Specifically, this model update strategy is established based on the temporal correlation of the learned filters and the fluctuation degree of the current frame response map, which updates the models only at high confidence levels and thus alleviates the model degradation problem. Meanwhile, in order to achieve robust tracking in both temporal and spatial domains, we introduce a temporal regularization term and extract background patches as samples for training. Extensive experiments on six aerial tracking datasets, i.e., UAVDT, DTB70, UAV123@10fps, UAVTrack112, UAV123, and VisDrone2018, demonstrate the effectiveness of the proposed tracker. Furthermore, we deploy the proposed tracker on quadrotor UAVs in real-world scenarios and successfully completed the UAV target tracking task. Our code and results are available at https://github.com/chenxlin222/ETEU .

The efficient convolution operator (ECO) have manifested predominant results in visual object tracking. However, in the pursuit of performance improvement, the computational burden of the tracker becomes heavy, and the importance of different feature layers is not considered. In this paper, we propose a self-adaptive mechanism for regulating the training process in the first frame. To overcome the over-fitting in the tracking process, we adopt the fuzzy model update strategy. Moreover, we weight different feature maps to enhance the tracker performance. Comprehensive experiments have conducted on the OTB-2013 dataset. When adopting our ideas to adjust our tracker, the self-adaptive mechanism can avoid unnecessary training iterations, and the fuzzy update strategy reduces one fifth tracking computation compared to the ECO. Within reduced computation, the tracker based our idea incurs less than 1% loss in AUC (area-undercurve).

Efficiently analyzing and classifying dynamically changing time series data remains a challenge. The main issue lies in the significant differences in feature distribution that occur between old and new datasets generated constantly due to varying degrees of concept drift, anomalous data, erroneous data, high noise, and other factors. Taking into account the need to balance accuracy and efficiency when the distribution of the dataset changes, we proposed a new robust, generalized incremental learning (IL) model ELM-KL-LSTM. Extreme learning machine (ELM) is used as a lightweight pre-processing model which is updated using the new designed evaluation metrics based on Kullback-Leibler (KL) divergence values to measure the difference in feature distribution within sliding windows. Finally, we implemented efficient processing and classification analysis of dynamically changing time series data based on ELM lightweight pre-processing model, model update strategy and long short-term memory networks (LSTM) classification model. We conducted extensive experiments and comparation analysis based on the proposed method and benchmark methods in several different real application scenarios. Experimental results show that, compared with the benchmark methods, the proposed method exhibits good robustness and generalization in a number of different real-world application scenarios, and can successfully perform model updates and efficient classification analysis of incremental data with varying degrees improvement of classification accuracy. This provides and extends a new means for efficient analysis of dynamically changing time-series data.

The typical industrial Internet of Things (IIoT) network system relies on a real-time data upload for timely processing. However, the incidence of device heterogeneity, high network latency, or a malicious central server during transmission has a propensity for privacy leakage or loss of model accuracy. Federated learning comes in handy, as the edge server requires less time and enables local data processing to reduce the delay to the data upload. It allows neighboring edge nodes to share data while maintaining data privacy and confidentiality. However, this can be challenged by a network disruption making edge nodes or sensors go offline or experience an alteration in the learning process, thereby exposing the already transmitted model to a malicious server that eavesdrops on the channel, intercepts the model in transit, and gleans the information, evading the privacy of the model within the network. To mitigate this effect, this paper proposes asynchronous privacy-preservation federated learning for mobile edge networks in the IIoT ecosystem (APPFL-MEN) that incorporates the iteration model design update strategy (IMDUS) scheme, enabling the edge server to share more real-time model updates with online nodes and less data sharing with offline nodes, without exposing the privacy of the data to a malicious node or a hack. In addition, it adopts a double-weight modification strategy during communication between the edge node and the edge server or gateway for an enhanced model training process. Furthermore, it allows a convergence boosting process, resulting in a less error-prone, secured global model. The performance evaluation with numerical results shows good accuracy, efficiency, and lower bandwidth usage by APPFL-MEN while preserving model privacy compared to state-of-the-art methods.

Network reciprocity has been successfully put forward (since M. A. Nowak and R. May's, 1992, influential paper) as the simplest mechanism-requiring no strategical complexity-supporting the evolution of cooperation in biological and socioeconomic systems. The mechanism is actually the network, which makes agents' interactions localized, while network reciprocity is the property of the underlying evolutionary process to favor cooperation in sparse rather than dense networks. In theoretical models, the property holds under imitative evolutionary processes, whereas cooperation disappears in any network if imitation is replaced by the more rational best-response rule of strategy update. In social experiments, network reciprocity has been observed, although the imitative behavior did not emerge. What did emerge is a form of conditional cooperation based on direct reciprocity-the propensity to cooperate with neighbors who previously cooperated. To resolve this inconsistency, network reciprocity has been recently shown in a model that rationally confronts the two main behaviors emerging in experiments-reciprocal cooperation and unconditional defection-with rationality introduced by extending the best-response rule to a multi-step predictive horizon. However, direct reciprocity was implemented in a non-standard way, by allowing cooperative agents to temporarily cut the interaction with defecting neighbors. Here, we make this result robust to the way cooperators reciprocate, by implementing direct reciprocity with the standard tit-for-tat strategy and deriving similar results.

With the increasingly severe environmental issues caused by fossil fuel consumption, clean energy technology represented by photovoltaics (PV) has attracted increasing research attention. However, the unknown parameter configuration of PV devices is related to conditions such as temperature and irradiance of the environment where the equipment is located, and existing methods often suffer from limited accuracy and reliability of parameter identification. To address these challenges, in this article, a Memory and Update Strategy-based Social Group Optimization (MUS-SGO) algorithm for PV parameter identification is proposed, aiming to enhance both accuracy and robustness. To strengthen local exploitation capability, a dynamic memory-guided strategy is employed. This strategy constructs a historical memory repository to store high-quality historical solutions and, together with the dynamic memory weights, guides MUS-SGO toward historical optimal regions. To maintain population diversity and accelerate convergence, an adaptive population update strategy is applied, which adaptively replaces a proportion of low-fitness individuals with new ones depending on the stage of MUS-SGO. Comparative experiments are conducted on the poly-crystalline KC200GT and mono-crystalline SM55 datasets under varying temperature and irradiance, using seven representative algorithms as baselines. The results demonstrate that MUS-SGO achieves a smaller root mean square error (RMSE) than the compared algorithms, with r 2 values close to 1.0. This indicates that MUS-SGO ensures both high accuracy and strong robustness for PV parameter identification.

In solving the portfolio optimization problem, the mean-semivariance (MSV) model is more complicated and time-consuming, and their relations are unbalanced because they conflict with each other due to return and risk. Therefore, in order to solve these existing problems, multi-strategy adaptive particle swarm optimization, namely APSO/DU, has been developed to solve the portfolio optimization problem. In the present study, a constraint factor is introduced to control velocity weight to reduce blindness in the search process. A dual-update (DU) strategy is based on new speed, and position update strategies are designed. In order to test and prove the effectiveness of the APSO/DU algorithm, test functions and a realistic MSV portfolio optimization problem are selected here. The results demonstrate that the APSO/DU algorithm has better convergence accuracy and speed and finds the least risky stock portfolio for the same level of return. Additionally, the results are closer to the global Pareto front (PF). The algorithm can provide valuable advice to investors and has good practical applications.

The strip crown directly affects the quality of strip steel. To enhance online crown control and product quality, an accurate and efficient strip crown prediction model is crucial. The accuracy of the prediction model is determined by the algorithm and dataset used in the modeling process. The accuracy can be enhanced by increasing algorithm complexity but it does not meet the requirements of online applications. Besides, the datasets collected for strip crown prediction are usually imbalanced, which impacts the modeling accuracy. In this paper, an enhanced strip crown prediction model based on KCGAN-ELM is established. A new hybrid algorithm KCGAN is proposed to deal with the imbalanced datasets. ELM is used to establish the strip crown prediction model. To meet the requirements of online applications, incremental learning is introduced, enabling the prediction model to update in real-time based on new production data. On this basis, an update strategy is devised to ensure the prediction model can maintain qualified prediction ability during the updating process. The experiment results indicate that the model trained with the dataset processed by KCGAN demonstrates a significantly enhanced prediction accuracy, achieving an RMSE of 2.86 μm and an R 2 value of 0.97. The proposed update strategy enhances the stability of prediction capability during the model updating process.

In the era of Industry 4.0, researchers in various fields have paid special attention to digital twin technology, which can realize real-time mapping between virtual and physical space. In this paper, a new general framework for response prediction of composite structures based on digital twins is proposed. The tensile testing process of standard samples of carbon fiber-reinforced composites (CFRCs) is used as the twinning object. Moreover, the development of a digital twin and composite structural response prediction based on the generic framework is demonstrated. First, standard CFRC tensile samples are prepared, and relevant raw data are acquired. Subsequently, the microscopic parameters of the standard CFRC tensile samples are obtained by scanning electron microscopy. Geometric measurements are performed to determine the macroscopic parameters, which, together with the material properties of carbon fibers and matrix, are used as the input parameters of a multi-scale virtual physical model (MVPM). The MVPM is used to simulate the actual tensile process using the multi-scale finite element method (FEM). Then, the real-time measurement data from the physical space are transferred to the virtual space through sensors. At the same time, the computationally time-consuming MVPM is downscaled to meet the real-time requirements for the online deployment of the digital twins. In this paper, the backpropagation (BP) neural network model is used to train the input and output parameter data of the MVPM to obtain a reduced-order model (ROM). In addition, to improve the prediction accuracy of the structural response of the digital twin, three model update strategies (MUS) of the ROM are proposed: 1) MUS 1 is based on the ROM, adding the tested sample historical data for the training model update strategy; 2) MUS 2 is based on the ROM 1, adding the measured real-time data of the current sample for training and updating to obtain the ROM 2; 3) MUS 3 is based on the predicted structural response data of ROM 2. Combined with the real-time measured data of the current sample, a higher-order fitting real-time correction is performed to obtain ROM 3. Finally, the tensile process of five CFRC standard samples is demonstrated based on the structural response prediction of the digital twin. The strain response prediction and contour visualization of the whole sample is achieved with limited strain gauge data. By comparison, MUS 2 has higher prediction accuracy than MUS 1 after adding the real-time measured data of the current sample. The prediction errors of MUS 1 and MUS 2 at the later stages of the stretching process are within 10%, with the minimum error of MUS 1 being 15.73% and that of MUS 2 being 3.36%. With the correction of high-order fitting, MUS 3 can achieve a stable prediction error of 20% or less in future moments, and the error can be reduced to less than 5%, reaching a minimum error of 0.44% at the critical tail section near tensile failure.

Snow ablation optimization (SAO) is a novel metaheuristic algorithm (MA). However, we observed certain issues in the original SAO, such as poor capacity in escaping from local optima and slow convergence. To address these limitations, we introduce two strategies: the asynchronous update strategy (AUS) and the top- k survival mechanism. We name our proposal SAO k -AUS. In the original SAO, the segregation of search and update delays the improved information sharing, and AUS integrates update processes following each individual’s search behavior, facilitating superior knowledge from elites. Additionally, the original SAO adopts an all-acceptance selection principle, maintaining diversity but cannot guarantee the solution quality. Thus, we introduce the top- k survival mechanism to ensure the survival of elites. Comprehensive numerical experiments on CEC2013 and CEC2020 benchmark functions, engineering problems, and image segmentation tasks were conducted to evaluate our proposal against eight state-of-the-art MAs. The experimental results and statistical analyses confirm the efficiency of SAO k -AUS. Moreover, the ablation experiments investigate the contribution of two strategies, and we recommend using both proposed strategies simultaneously. The source code of this research is made available in https://github.com/RuiZhong961230/SAO_k-AUS .