
Effective communication among autonomous agents is crucial for coordination and solving complex tasks within multi-agent systems. To formalize interactions between agents, social accessibility relations are often utilized. Current research employs model checking algorithms to verificate social commitment properties in multi-agent systems. In fuzzy multi-agent systems, direct quantification and computation of commitment attributes pose challenges. This paper introduces an indirect fuzzy model checking algorithm designed to convert social commitments in uncertain scenarios into quantifiable attributes for verification. Firstly, we propose a fuzzy communicative interpreted system model to represent multi-agent systems with uncertain communication. We then improve fuzzy computation tree logic by adding modalities for commitments and fulfillment, resulting in a fuzzy computation tree logic with commitments for describing system properties related to commitments. A fuzzy model checking algorithm is subsequently presented. This algorithm converts the task of model checking fuzzy computation tree logic with commitments based on fuzzy interpreted systems into model checking fuzzy computation tree logic based on fuzzy Kripke structures. We conclude by providing proofs of correctness and complexity analysis of our algorithm. Furthermore, we demonstrate the effectiveness of our approach for model checking social commitments under fuzzy conditions through simulation experiments on an online shopping system.
Wireless sensing combined with deep learning has become a research hotspot in recent years, but most of the existing work is limited by the sensing distance. The establishment of several sensing models in recent years has proven the long-distance sensing ability of LoRa devices, but there is still insufficient work on how to combine LoRa devices with deep learning. This paper analyzes the propagation and variation laws of LoRa signals affected by human activities, proposes a LoRa signal processing method to extract features suitable for deep learning networks. By comparing current advanced models with commonly used signal processing methods, the final result is good. For the types and roles of activities in one room, the accuracy exceeds 94%. Compared with common signal processing methods, it also saves space and time resources, which proves the effectiveness of the proposed method.
Since multi-person pose estimation in crowded scenes still suffers from problems such as small detection targets, resulting in low pose estimation accuracy, we propose a correction algorithm for multi-person pose estimation based on the improved YOLOv5. Firstly, in the backbone network of YOLOv5, the Jumping Attention Mechanism module is incorporated to help the networks find the region of interest in the images; Secondly, in the neck network, the joint use of the Jump Attention Mechanism module and the Transformer encoder allows the network to acquire global information and contextual information. Finally, the key point object information obtained from the network prediction is used to correct the pose object information to get the final multi-person pose estimation results. The experimental results show that the method in this paper improves AP50 by 2.2% and AP75 by 3.3% over YOLOv5 on the COCO dataset, which verifies the accuracy and robustness of the approach in this paper.
The maximum mean discrepancy (MMD) as a representative distribution metric between source domain and target domain has been widely applied in unsupervised domain adaptation (UDA), where both domains follow different distributions, and the labels from source domain are merely available. However, MMD and its class-wise variants possibly ignore the intra-class compactness, thus canceling out discriminability of feature representation. In this paper, we endeavor to improve the discriminative ability of MMD from two aspects: 1) we re-design the weights for MMD in order to align the distribution of relatively hard classes across domains; 2) we explore a focally contrastive loss to trade-off the positive sample pairs and negative ones for better discrimination. The intergration of both losses makes the intra-class features close as well as push away the inter-class features far from each other. Moreover, the improved loss is simple yet effective. Our model shows state-of-the-art compared to the most domain adaptation methods.
在网络信息快速发展的时代背景下,不同的音乐爱好者对音乐信息检索的需求也在不断提高,音乐分类成为了一个非常值得研究的问题.提出了将耗散均匀搜索粒子群优化算法 DUPSO 与基于距离排序的支持向量预选取算法 DSVM 相结合的 DUPSO-DSVM 民歌快速分类算法.该算法利用DUPSO 算法对SVM的惩罚系数C和核函数参数g 进行优化,并利用DSVM算法优化 DUPSO 算法的参数优化时间.实验结果表明,在使用了 DUPSO-DSVM 算法之后,算法的训练时间只占未使用DUPSO-DSVM算法的 26.26%,民歌分类正确率为 84%,但仍然保持着较高的分类正确率.