Federated semi-supervised learning (FSSL) has emerged as a promising paradigm to address the challenges of data privacy and the scarcity of labeled data. Most existing algorithms adapt the pseudo-labeling and consiste...
详细信息
ISBN:
(数字)9798350380323
ISBN:
(纸本)9798350380330
Federated semi-supervised learning (FSSL) has emerged as a promising paradigm to address the challenges of data privacy and the scarcity of labeled data. Most existing algorithms adapt the pseudo-labeling and consistency regularization methods in traditional semi-supervised learning (SSL) to exploit raw data from unlabeled clients. However, these methods always use the same high fixed threshold to filter out noisy pseudo-labels for each class, without considering the varying learning conditions and difficulty levels associated with different classes, resulting in a decline in the performance of global model, especially under non-Independent and Identically Distributed (Non-IID) settings. To this end, we propose FedTMatch, a learning framework which adjusts the threshold for each class based on the learning status of each client. Furthermore, we introduce a self-supervised rotation loss in unlabeled clients to increase the utilization of raw data and correct the feature space formed by pseudo-label data to weaken the impact of Non-IID data distribution on global model's performance. Extensive experiments on three benchmark datasets demonstrate the effectiveness of our approach.
Muzzle speed and efficiency are two important factors that constrain the development of reluctance coil launchers. In order to improve the muzzle speed of the reluctance coil launcher, the number of stages of the laun...
详细信息
ISBN:
(数字)9789038661353
ISBN:
(纸本)9798350352733
Muzzle speed and efficiency are two important factors that constrain the development of reluctance coil launchers. In order to improve the muzzle speed of the reluctance coil launcher, the number of stages of the launcher is bound to increase, which leads to the complication of the simulation model, long simulation time consuming and low optimisation efficiency. Aiming at the above problems, this paper proposes a fast optimisation method for high-speed reluctance coil launchers based on genetic algorithm and machine learning. Firstly, based on a 7-stage reluctance coil launcher, determine that the wire diameter and the number of coil layers are the optimization objects; based on machine learning, the wire diameter and the number of coil layers are the optimization objects. Based on machine learning, the fast calculation model is trained withthe wire diameter and the number of coil layers as inputs and the muzzle speed as output; the fast calculation model is invoked withthe genetic algorithm, and the maximum muzzle speed of the launcher is taken as the fitness function, so as to achieve the fast optimisation of the parameters of the high-speed reluctance coil launcher stage by stage. the results show that after optimisation, an muzzle speed of 113.59m/s can be achieved under the condition of 32-stage launcher with an efficiency of 9.96%.
In multi-site brain disease diagnosis studies, traditional centralized training methods necessitate sharing medical data, posing significant privacy risks. Federated learning (FL) offers a privacy-preserving solution ...
详细信息
ISBN:
(数字)9798350380323
ISBN:
(纸本)9798350380330
In multi-site brain disease diagnosis studies, traditional centralized training methods necessitate sharing medical data, posing significant privacy risks. Federated learning (FL) offers a privacy-preserving solution by enabling global model training through aggregating locally trained models from multiple data centers without sharing raw data. However, current FL approaches rely on a server-based network topology, where central server failure disrupts training. Additionally, data heterogeneity across sites often slows convergence and reduces accuracy. To overcome these issues, we introduce a decentralized personalized federated learning collaborative aggregation network (pFedCAN). this framework has two core components: (1) separating local models into shared and personalized layers, and (2) forming a collaborative aggregation network via similarity detection in the shared layers. Specifically, each center trains its local model, then separates it into shared and personalized layers. the shared layer is exchanged with other centers, while the personalized layer remains local. Data centers analyze similarities in received shared layers to build a collaborative network, where shared layers from similar centers are aggregated to refine the model. this approach flexibly adapts to varying levels of data heterogeneity, enhancing model training efficiency. Validation on public datasets, ABIDE I and ADHD, shows that the proposed method outperforms current leading techniques.
Partial shading (PS) significantly impacts the performance of photovoltaic (PV) systems, leading to reduced efficiency. Traditional mitigation methods involve bypass diodes or PV module reconfiguration. Accordingly, t...
详细信息
ISBN:
(数字)9798350350708
ISBN:
(纸本)9798350350715
Partial shading (PS) significantly impacts the performance of photovoltaic (PV) systems, leading to reduced efficiency. Traditional mitigation methods involve bypass diodes or PV module reconfiguration. Accordingly, the highest and lowest layer-based exchange (HLLBE) algorithm was proposed. the HLLBE system employs a Switching Block Array (SBA) for flexible reconfiguration without rewiring. By sorting and swapping irradiance readings, the HLLBE algorithm achieves uniform current distribution across PV array layers. Unlike static configurations, HLLBE dynamically adapts to shade patterns, offering an efficient solution. Simulations on a 3 × 3 PV array demonstrated HLLBE's effectiveness in enhancing output power and reducing multiple peaks compared to conventional total-cross-tied (TCT) configurations. Additionally, HLLBE minimizes operating switches through a switching matrix during dynamic reconfiguration (DR).
the rapid development of artificial intelligence technology has drawn widespread attention to the problem of target recognition in open environments. Traditional target recognition methods are easily affected by envir...
详细信息
ISBN:
(数字)9798350355642
ISBN:
(纸本)9798350355659
the rapid development of artificial intelligence technology has drawn widespread attention to the problem of target recognition in open environments. Traditional target recognition methods are easily affected by environmental changes and interference noise, leading to insufficient recognition accuracy and robustness. this paper proposes a robust target recognition method based on causal inference, aiming to improve the robust performance of image recognition models in open environments. the method first extracts features from images, analyzes the impact of noise features such as the environment in image features, and derives the essential feature properties of the target. then, the causal relationship is introduced into the target recognition, and by analyzing the relationship between the environment, the target, and the observation data, the features of the target are combined withthe observation data to achieve accurate and stable recognition of the target. the proposed method has achieved good target recognition effects on multiple open datasets, and the experimental results show that the algorithm can effectively solve the problem of poor robustness of the target recognition model in open environments, providing strong support for practical applications. through the application of this method, the stability and robustness of the target recognition model can be significantly improved, which has important practical application value.
the tracking control of an unmanned autonomous vehicle (UA V) is one of the key factors determining the performance and effectiveness of autonomous navigation. A controller based on Linear Parameter Varying - Model Pr...
详细信息
ISBN:
(数字)9798350350302
ISBN:
(纸本)9798350350319
the tracking control of an unmanned autonomous vehicle (UA V) is one of the key factors determining the performance and effectiveness of autonomous navigation. A controller based on Linear Parameter Varying - Model Predictive Control (LPV - MPC) is designed to track the desired trajectory of the UAV. the Linear Parameter Varying (LPV) modeling and control strategy allows the representation of the UAV nonlinear complex model in a linear-like manner which can be used in a linear Model Predictive Control (MPC). the main advantage of the LPV -MPC technique is its ability to improve tracking solutions very close to nonlinear MPC but with significantly low computational costs, enabling LPV-MPC to be used for real-time operations. the simulation results of this paper show that the designed LPV-MPC performs well in tracking the desired trajectory through each waypoint. Furthermore, experiments with different numerical integration methods have been investigated, especially Euler's method, and the Runge-Kutta method. the results showed that the Runge-Kutta method performed better with smaller horizon periods, with minimum tracking errors and shorter elapsed time.
the perception and transportation of soft objects are critical tasks in daily life, and recent sensor advancements enable robots to safely perform these tasks in human-shared environments. this study advances elastic ...
详细信息
ISBN:
(数字)9798331518301
ISBN:
(纸本)9798331518318
the perception and transportation of soft objects are critical tasks in daily life, and recent sensor advancements enable robots to safely perform these tasks in human-shared environments. this study advances elastic linear object transportation by introducing a collaborative framework where a human leads, providing task-specific instructions to guide the robot. Unlike conventional methods focused on fixed object manipulation, our approach dynamically controls and converges feature points of the elastic object into desired configurations during transportation. We developed an online model estimation technique using a least-squares optimization algorithm, with an exponential forgetting mechanism to adaptively update the model under disturbances from sensors and human interaction. Validated in experiments with a 6-DOF robot and depth camera, our method demonstrated robustness across varied scenarios, achieving faster convergence and reduced positional fluctuation compared to conventional gradient descent approaches.
We address an optimal reachability problem for a planar manipulator in a constrained environment. After introducing the optmization problem in full generality, we practically embed the geometry of the workspace in the...
详细信息
Large Language models (LLMs) have demonstrated significant effectiveness across various NLP tasks, including text ranking. this study assesses the performance of large language models (LLMs) in listwise reranking for ...
详细信息
ISBN:
(数字)9798331531225
ISBN:
(纸本)9798331531232
Large Language models (LLMs) have demonstrated significant effectiveness across various NLP tasks, including text ranking. this study assesses the performance of large language models (LLMs) in listwise reranking for limited-resource African languages. We compare proprietary models RankGPT3.5, Rank4o-mini, RankGPTo1-mini and RankClaude-sonnet in cross-lingual contexts. Results indicate that these LLMs significantly outperform traditional baseline methods such as BM25-DT in most evaluation metrics, particularly in nDCG@10 and MRR@100. these findings highlight the potential of LLMs in enhancing reranking tasks for low-resource languages and offer insights into cost-effective solutions.
Airport Traffic Control Tower (ATCT) is one of the most important facilities at civil aviation airports. the location and height of ATCT will directly affect the controller's observation and judgment. In order to ...
详细信息
ISBN:
(数字)9798350380323
ISBN:
(纸本)9798350380330
Airport Traffic Control Tower (ATCT) is one of the most important facilities at civil aviation airports. the location and height of ATCT will directly affect the controller's observation and judgment. In order to evaluate and compare the visibility capability of different tower construction schemes, a comprehensive evaluation method is set up. this method includes building an index system of obligatory and preference indicators, forming calculation methods for each indicator, and establishing a comprehensive evaluation model. In the evaluation model, two networks have been constructed based on Coefficient Strategy (CS) and Level Strategy (LS) respectively. Selecting a certain airport in East China as a case study, CS-BP and LS-BP network models of 5-10-1 in topology have been set up after 12000 typical samples for training respectively. the research results show that LS-BP network has the smaller Mean Squared Error (MSE) and greater regression coefficient compared to CS-BP network. And the results of CS-BP were similar to Traditional Strategy (TS). therefore, selecting LS-BP network to evaluate calculated indicators of given schemes and considering that scheme 3 was best. the entire method can serve as a reference for decision-making by air traffic control unit.
暂无评论