The problems of RFID indoor positioning algorithm LANDMARC are optimized. In view of the abnormal correlation degree of local reference tags in this algorithm, this paper repairs the correlation degree of bad tags. Aiming at the problem of label misselection in the process of selecting adjacent labels, an improvement of minimum target area block was proposed to reduce the influence of bad labels, and then the location of target labels was estimated. Simulation results show that the proposed optimization scheme is significantly improved compared with the original algorithm, the positioning accuracy is increased by 22%, and the maximum error is reduced by 27%.
Edge computing is to migrate terminal tasks to the edge server for execution,and reduce task execution delay and terminal energy consumption through reasonable migration decision.In view of the fact that a large number of literatures only consider the unloading decision of multi-task single server and ignore the balance of computing tasks between servers,a multi-task multi-server joint optimization strategy based on particle swarm optimization(PSO)is proposed.Taking the delay and energy consumption as the optimization objectives,considering the queuing delay and the balance of the amount of computation of the edge computing server(MEC),the global computing resources were fully utilized,which further reduced energy consumption and time delay.Simulation results show that,compared with other algorithms,the average task cost of the proposed algorithm is the smallest,which proves the feasibility and high efficiency of the method.
Identity authentication is a very important aspect of information security theory. Usually, the problem that identity authentication is prone to is that the user name and password are transmitted in the network in plaintext. The security is not enough, and it is very vulnerable to replay attacks and dictionary attacks. Dynamic passwords Although authentication is more secure, if there is a problem with the time synchronization between the server and the client, the user will not be able to login for a long time. Moreover, the password of the dynamic password is the same within a period of time, and it is also vulnerable to replay attacks. Therefore, a decentralized dynamic identity authentication system based on blockchain isn’t proposed. By using a public key + nonce (integer number that increases from 0) to sign in, It only solves the problem of over-centralized identity authentication, but also solves the replay attack. Improved the security of the identity authentication process.
Aimed at the problem of gradient disappearance and network degradation caused by the use of recurrent neural networks in traditional question answering models,an automatic question answering model based on IMGRU-Seq2seq(Identity Mapping Gated Recurrent Unit-Sequence to Sequence)is proposed.The text was represented by weighted word vectors through the TF-IDF method.Based on the gated recurrent unit,the batch normalization technology and the rectified linear unit activation function were combined and the identity mapping was added to construct the IMGRU model.As the semantic extraction unit of the question answering model,the bidirectional IMGRU introduced the attention mechanism and the beam search algorithm to realize automatic question and answer.The experimental results show that the proposed method is 18.87％and 4.35％higher than the existing methods BLEU and ROUGE-L respec-tively.
Autoencoders have been successfully used for graph embedding, and many variants have been proven to effectively express graph data and conduct graph analysis in low-dimensional space. However, previous methods ignore the structure and properties of the reconstructed graph, or they do not consider the potential data distribution in the graph, which typically leads to unsatisfactory graph embedding performance. In this paper, we propose the adversarial attention variational graph autoencoder (AAVGA), which is a novel framework that incorporates attention networks into the encoder part and uses an adversarial mechanism in embedded training. The encoder involves node neighbors in the representation of nodes by stacking attention layers, which can further improve the graph embedding performance of the encoder. At the same time, due to the adversarial mechanism, the distribution of the potential features that are generated by the encoder are closer to the actual distribution of the original graph data; thus, the decoder generates a graph that is closer to the original graph. Experimental results prove that AAVGA performs competitively with state-of-the-art popular graph encoders on three citation datasets.
In recent years, due to the rise of online social platforms, social networks have more and more influence on our daily life, and social recommendation system has become one of the important research directions of recommendation system research. Because the graph structure in social networks and graph neural networks has strong representation capabilities, the application of graph neural networks in social recommendation systems has become more and more extensive, and it has also shown good results. Although graph neural networks have been successfully applied in social recommendation systems, their performance may still be limited in practical applications. The main reason is that they can only take advantage of pairs of user relations but cannot capture the higher-order relations between users. We propose a model that applies the hypergraph attention network to the social recommendation system (HASRE) to solve this problem. Specifically, we take the hypergraph’s ability to model high-order relations to capture high-order relations between users. However, because the influence of the users’ friends is different, we use the graph attention mechanism to capture the users’ attention to different friends and adaptively model selection information for the user. In order to verify the performance of the recommendation system, this paper carries out analysis experiments on three data sets related to the recommendation system. The experimental results show that HASRE outperforms the state-of-the-art method and can effectively improve the accuracy of recommendation.
Extraction of livers and tumors from CT(Computed Tomography) volumes is a challenging task. In recent years, with the development of deep learning technology, The U-Net architecture is widely used in medical image segmentation. We propose a residual network-based and efficient multi-scale attention method EPA-UNet.In the whole network, the U-Net framework is selected as the baseline, and the context-aware residual block feature encoder decoder enhances the feature extraction capability and solves the deep network degradation problem. An Efficient Channel Attention (ECA) is introduced in the residual unit, which uses an attention mechanism to assign weights to the input sequence and can suppress irrelevant image background features. In order to solve the difficult optimization problem of merging objects of different sizes in the skip connection of U-Net, Attention Gates (AG) are introduced into the skip connection for optimization, and useful features are further learned to improve the sensitivity and accuracy of the module. Second, Atrous Spatial Pyramid Pooling (ASPP) is used to replace the bottleneck layer to acquire multi-scale image information through different receptive fields. Our structure was evaluated on the LiTs dataset while using three commonly used segmentation metrics which are the Dice coefficient, 95HD, and ASD. The method obtained 96.6% Dice coefficient for liver segmentation and 82.5% Dice coefficient for tumor segmentation. The highest accuracy was achieved for all three metrics compared to other methods. Our experiments used a five-fold cross-validation method, which effectively reduces the random errors in the model evaluation and improves the accuracy of the evaluation results.
With the increasing size of code and the increasing complexity of code files,code defect detection tools need to adopt parallel scheduling method for scheduling.In order to better use parallel method for scheduling and improve the efficiency of defect detection and utilization of hardware resources,we propose a method to predict the cost of the module tested in code defect detection.According to the characteristics of the defect testing system(DTS)defect detection process,the time cost feature and space cost feature were extracted.The semantic feature was extracted by deep memory network.The time cost feature and semantic feature were fused to get the fusion feature,and the regression model was used to predict the time cost of the fusion feature and the space cost of the space cost feature.Experimental results on 8 open source C projects show that the proposed method has a good performance in cost prediction.
Solid-state transformer (SST) is the core equipment of energy router in Energy Internet. As distributed generation units are often in multivariable, strong coupling and violent unknown disturbance, these seriously affect the stable operation of the system. To solve this problem, a sliding mode active disturbance rejection controller (SMADRC) is designed for solid-state transformer rectifier stage. The state variables observed by the linear extended state observer (LESO) are used for feedforward compensation adjustment in the sliding mode controller (SMC) to reduces random fluctuation of bus voltage caused by disturbance. The system is verified by the simulation platform, and the results show that the dynamic performance and grid-connection quality of the system are significantly improved by the sliding mode ADRC.
The development of high-throughput sequencing technology provides an opportunity to obtain multi-omics data for liver cancer,However,omics data often comes from different platforms and has different attributes, it has the characteristics of high feature dimension and small sample size. This will increase the overfitting of the model and the imbalance of categories,and the cross-platform integration analysis of omics data will challenge the traditional data analysis methods. In this regard, the Hierarchical Integrated Stacked Encoder (HI-SAE) is proposed.which can achieve deeper feature learning and data integration while reducing the differences caused by the characteristics of the data itself. Finally,the integrated feature expression is used to identify the subtype of liver cancer by softmax classifier. Experiments show that the classification accuracy when using Hi-SAE method for feature learning is 3.7% higher than that when using PCA, and 7.6% higher than that when using NMF.
In recent years, Generative Adversarial Network (GAN) has quickly become the most popular deep generative model framework, and it is also the most popular topic in the current deep learning research field. Although the generative adversarial network has achieved remarkable results from text description to image generation, when a complex image containing multiple objects, the position of each object will be blurred and overlapped, and the edges of the generated image will be blurred and local textures will be unclear. Usually given text description can generate the corresponding rough image, but there are still some problems in the image details. In order to solve the above problems, on the basis of Stack GAN, a scene graph-based stacked generative confrontation network model (Scene graph stack GAN, SGS-GAN) is proposed, which converts the text description into The scene graph uses the scene graph as the condition vector and inputs the random noise into the generator model to obtain the result image. The experimental results show that the Inception store of the SGS-GAN model on the Visual Genome and COCO data sets reached 6.64 and 6.52, respectively, which were increased by 0.212 and 0.219 compared to Sg2Im. This proves that the diversity and vividness of the generated samples and the sharpness of the image are obviously improved after the number of times of training and the input of the scene graph.
近年来基于物流机器人的智能物流系统逐渐应用于仓库运营中,大幅度提高了订单拣选的效率,减少了人力劳动.为解决电商行业广泛应用的订单拣选问题,以最小化完成所有订单拣选所需的货架搬运次数为目标,建立整数规划模型,并设计订单拣选排程算法.由于该问题属于NP-hard问题,从订单排序和货架排序两个角度,分别设计两种变邻域搜索算法VNS-OS和VNS-RS,进而设计交替求解订单和货架排序的AH算法.在小规模算例下,对比商业求解器CPLEX在指定时间内求得的解,证明了该算法在计算时间和目标函数上的优势.在大规模算例下,验证了两种VNS算法对初始解的优化效果明显,平均优化程度高于10%,而AH算法在计算时间上具有很大优势.利用所提出的算法对储位分配策略的影响进行研究,为基于机器人的仓库运营提供参考.
数据传输是微控制器运行中必不可缺少的操作,CPU绝大部分时间也都花费在数据的搬运上.直接存储器访问(Direct Memory Access,DMA)是大多数处理器中提高数据传输速度的方式之一.在分析了多种微控制器芯片中DMA控制器工作原理的基础上,总结出其共性技术,从而提出一种基于STM32L431RC芯片的基础构件封装方式,并将其与UART串口相结合应用到实践中.实践证明,DMA具有传输速度快、准确率高等优点,同时也提高了CPU的工作效率.
事件机制是mbedOS中一种对任务进行实时控制的同步手段.为了深入分析事件机制响应调度的实时性,提出一种基于printf函数的时序分析方法,通过中断与线程的同步实验,在简要分析了响应调度的理论时间和过程的基础上,对实际响应时间进行了测试和分段解析,进一步探究了影响响应时间的要素.实验结果表明,该方法能有效完成实时性分析,提高mbedOS中事件机制对任务的更精确控制,也为其他RTOS(Real-Time Oper-ating System)的实时性分析提供一定的技术基础.
网络入侵检测系统是网络安全的重要组成部分,被用来检测并防御各种外来攻击.作为入侵检测尤为重要的方法之一,传统的异常检测模型存在误报率高、性能差等问题.针对这些问题,提出一种网络流量异常检测模型Enhanced BiGAN(EB-GAN).使用BiGAN基本网络框架学习真实复杂数据的分布;GAN的鉴别器使用双向长短期记忆网络捕捉序列分布的时间相关性,添加注意力机制使模型更有效地找到输入数据与当前输出数据之间的相关信息,从而提高检测率;结合WGAN和混合增强GAN的损失函数使得模型训练收敛更快.利用KDDCUP1999 数据集对该方法进行了实验评价.实验结果表明,该方法优于对比的流量异常检测方法.
用户的行为序列既包含了用户的短期兴趣,也包含了用户的长期偏好.针对此类问题,提出一个基于用户长短期兴趣的自注意力模型.使用循环神经网络来处理变长的用户序列,得到用户的长期兴趣表示;用自注意力网络计算序列中的每个项目对预测用户下一次交互项目的重要性程度,得到用户的短期兴趣表示;将循环神经网络的输出作为查询输入到自注意力网络中,得到综合长短期兴趣的用户表示,并用这个表示来预测用户的下一次交互.该算法在三个真实世界的数据集上评估了提出的模型,其中命中率指标提高7%～30%.
ICS接入互联网后,原有的物理隔离被打破,导致ICS固有漏洞和缺陷被暴露在互联网中,极易受到入侵者的攻击,已经严重威胁工业生产的安全.为了保证工业生产的安全稳定,大量研究人员针对ICS入侵检测技术进行研究.根据ICS独特性和高周期性将现有的研究方法分为基于特征的入侵检测方法和基于行为周期的入侵检测方法,并详细分析了两种方法的差异和优缺点,最后对IDS研究方向进行展望.
EAST(Experimental Advanced Superconducting Tokamak)等离子体垂直位移快速控制系统在保障等离子体平稳运行、获得高性能等离子体方面起到重要的作用.目前EAST垂直位移控制算法集成在等离子体控制系统PCS(Plasma Control System)中,利用反射内存卡将控制命令传输到快控电源从而进行控制,这种方案具有较长的系统间通信延迟以及输出读写延迟.基于PCS中控制算法,设计一套独立的垂直位移控制算法,使得系统从PCS中剥离,减少系统通信延迟,并提出一种算法模拟方案,利用历史数据实现算法模拟.同时引入一种新型输出机箱替代原有的输出方式,通过光纤传输,缩短输出延迟,提高系统响应速度,满足控制需求.
在对Prophet路由协议研究的基础上,利用节点相遇间隔时间具有指数分布的特性提出一种结合节点剩余相遇间隔改进 Prophet 路由算法-PRREI(Prophet routing with residual encounter interval of nodes).该算法在选择中继节点时,充分考虑了相遇节点与目的节点到下次相遇时,在剩余相遇时间内相遇的概率,使得相遇概率值估算既有动态性又有时效性.此外,引入了动态缓冲区管理策略,利用消息的剩余生存时间对消息的重要性进行度量来确定消息转发和删除的优先级,避免网络拥塞.结果表明,该算法能有效提高投递率,降低了时延和传输开销.
