Network

摘要： Neuroimaging techniques such as magnetic resonance imaging(MRI)and positron emission tomography provide unique in vivo data to analyze structural and functional connectivity of the whole brain.Recent advances in small animal neuroimaging have opened new opportunities for the study of structure-function interactions in healthy and diseased brain networks,which are essential to develop therapies targeting network reorganization associated with functional improvement.

摘要： In recent years,the construction of network security has been highly valued at the national level,and many laws and regulations such as Cybersecurity Law,Data Security Law,Personal Information Protection Law have been successively introduced,which means strong policy support for the development of China’s network security industry.According to the data from iiMedia Research,in 2021,the scale of China’s network information security market reached RMB 92.68 billion.With the continuous implementation of policy stimulus and technological progress,the growth scale of China’s network information security market will further expand in the future,and it is estimated that the market scale will exceed RMB 140 billion in 2023.

摘要： Effective thermal conductivity of soils can be enhanced to achieve higher efficiencies in the operation of shallow geothermal systems.Soil cementation is a ground improvement technique that can increase the interparticle contact area,leading to a high effective thermal conductivity.However,cementation may occur at different locations in the soil matrix,i.e.interparticle contacts,evenly or unevenly around particles,in the pore space or a combination of these.The topology of cementation at the particle scale and its influence on soil response have not been studied in detail to date.Additionally,soils are made of particles with different shapes,but the impact of particle shape on the cementation and the resulting change of effective thermal conductivity require further research.In this work,three kinds of sands with different particle shapes were selected and cementation was formed either evenly around the particles,or along the direction parallel or perpendicular to that of heat transfer.The effective thermal conductivity of each sample was computed using a thermal conductance network model.Results show that dry sand with more irregular particle shape and cemented along the heat transfer direction will lead to a more efficient thermal enhancement of the soil,i.e.a comparatively higher soil effective thermal conductivity.

摘要： Domestication and genetic improvement of maize improve yield and stress tolerance due to changes in morphological and physiological properties, which likely alter rhizosphere microbial diversity. Understanding how the evolution of maize germplasm impacts its rhizobacterial traits during the growth stage is important for optimizing plant-microbe associations and obtaining yield gain in domesticated germplasms. In this study, a total of nine accessions representing domestication and subsequent genetic improvement were selected. We then sequenced the plant DNA and rhizobacterial DNA of teosinte, landraces and inbred lines at the seedling, flowering and maturity stages in a field trial. Moreover, the soil chemical properties were determined at the respective stages to explore the associations of soil characteristics with bacterial community structures. The results showed that domestication and genetic improvement increased the rhizobacterial diversity and substantially altered the rhizobacterial community composition. The core microbiome in the rhizosphere differed among germplasm groups. The co-occurrence network analysis demonstrated that the modularity in the bacterial network of the inbred lines was greater than those of teosinte and the landraces. In conclusion, the increased diversity of the rhizobacterial community with domestication and genetic improvement may improve maize resilience to biotic stresses and soil nutrient availability to plants.

摘要： Sentence simplification is an essential task in natural language processing and aims to simplify complex sentences while retaining their primary meanings.To date,the main research works on sentence simplification models have been based on sequence-to-sequence(Seq2Seq)models.However,these Seq2Seq models are incapable of analysing the hierarchical structure of sentences,which is of great significance for sentence simplification.The problem can be addressed with an ON-MULTI-STAGE model constructed based on the improved MULTI-STAGE encoder model.In this model,an ordered neurons network is introduced and can provide sentence-level structural infor-mation for the encoder and decoder.A weak attention connection method is then employed to make the decoder use the sentence-level structural details.Experimental results on two open data sets demonstrated that the constructed model outperforms the state-of-the-art baseline models in sentence simplification.

摘要： Should the article be accepted and published by Agricultural Science & Technology, the author hereby grants exclusively to the editorial department of Agricultural Science & Technology the digital reproduction, distribution, compilation and information network transmission rights.

摘要： China Telecom is writing a new chapter on the journey of enhancing China’s strength in cyberspace,building a digital China as well as maintaining cyber security.STAYING true to its mission of building the world’s largest 5G network through co-construction and sharing.

摘要： Brassica rapa(AA,2n=20)and Raphanus sativus(RR,2n=18)are popular vegetables containing multiple secondary metabolites,such as flavonoids and anthocyanins.Nevertheless,fewstudies on miRNA expression profiles and potential regulatory network of flavonoid metabolism have been reported in the intergeneric hybrid of B.rapa and R.sativus(Brassicacoraphanus,AARR,2n=38).Here,themiRNAs of Brassicacoraphanus and its parents were first identified by small RNA sequencing.It was revealed that 315 conserved miRNAs and 193 novel miRNAs were found in Brassicacoraphanus and its parents.Moreover,the miRNA expression profiles were characterized and 204 miRNAs with differential expression were obtained.Further analysis showed that 184 miRNAs were non-additively expressed in Brassicacoraphanus.At last,the potential regulatory network of flavonoid metabolism in Brassicacoraphanus was analyzed by multi-omics data,suggesting possible roles of miR858a and miR157a in regulating flavonoid synthesis pathway in Brassicacoraphanus.Taken together,this study provides useful resources for probing globalmiRNA expression patterns in Brassicacoraphanus,which facilitates regulatory mechanism research on the biosynthesis pathway of secondary metabolites in Brassicacoraphanus.

摘要： Guest editorial The emerging applications,suchas Augmented and Virtual Realities(AR/VR),InternetofThings(IoT),4K/8Kstreaming,raisestrongrequirementsto movecomputationfrom thecloudtotheedgestobecloser tousers.There are tremendous possibilities for the network edge,which may includeavariety ofentities,such as small datacenters,end devices,and resource-abundant network nodes.These together provide the network computation and intelligence to users.

摘要： In the development of technology in various fields like big data analysis,data mining,big data,cloud computing,and blockchain technology,security become more constrained.Blockchain is used in providing security by encrypting the sharing of information.Blockchain is applied in the peerto-peer(P2P)network and it has a decentralized ledger.Providing security against unauthorized breaches in the distributed network is required.To detect unauthorized breaches,there are numerous techniques were developed and those techniques are inefficient and have poor data integrity.Hence,a novel technique needs to be implemented to tackle the new breaches in the distributed network.This paper,proposed a hybrid technique of two fish with a ripple consensus algorithm(TF-RC).To improve the detection time and security,this paper uses efficient transmission of data in the distributed network.The experimental analysis of TF-RC by using the metric measures of performance in terms of latency,throughput,energy efficiency and it produced better performance.

摘要： 本发明公开了一种用于加速Deep Q‑Network算法的硬件架构及其设计空间探索方法。硬件架构包括：通用处理器模块负责与外部环境进行交互和实现奖励函数的计算，也负责Deep Q‑Network算法经验池的维护；外部DDR存储器负责Deep Q‑Network算法的经验池的存储；AXI总线接口为通用AXI总线接口结构，负责实现通用处理器与FPGA可编程逻辑模块之间控制信号和数据信号的传递与反馈；Target Q模块负责实现Target Q网络的前向推理计算；Current Q模块负责实现Current Q网络的前向推理和反向传播。本发明在高度优化FPGA硬件架构下，实现Deep Q‑Network算法的实时计算。

摘要： 本发明公开了基于Relation‑Network算法的生态生物识别方法，包括如下步骤：发起生态生物识别请求，根据请求在生态环境中采集多个生物图像；对采集的生物图像数据进行优化求精，得到优化后的生物图像；收集历史生态生物图像，将收集的历史生态生物图像进行分类，并建立生物图像数据库；通过Relation‑Network算法对采集的生物图像和生物图像数据库中的样本进行对比匹配；通过生物图像和样本之间的匹配度进行生态生物的识别。本发明通过设置Relation‑Network算法，通过构建神经网络来计算两个样本之间的距离从而分析匹配程度，能够有效的提高生物的识别率。

摘要： 本发明公开了一种基于Echo State Network的图像分类方法，包括构建网络结构、训练中的数据传输和更新、输出训练和图像分类处理的步骤，其中，储备池的各个参数设置为：储备池光谱半径λ为1.25，储备池大小Nx为2000，储备池输入单元的尺度IS为1，储备池的稀疏度SD为1e‑8，泄漏率α为0.15，空转图像数量占10％。本发明提供的基于Echo State Network的图像分类方法，通过使用回声状态网络(ESN)进行图像的分类处理，解决了卷积网络在使用上复杂度高、速度较慢的问题。

摘要： 本发明公开了一种基于Deep Q‑Network改进的俄罗斯方块智能Agent学习方法，包括以下步骤：1)获取当前游戏状态下方块的统计特征，根据游戏画面采集游戏画面每帧的统计特征序列；2)通过深度神经网络来进行俄罗斯方块智能Agent Q值的估计，所述Q值为俄罗斯方块agent在采取某个action后的表现对应的奖励分数；3)在Q值估计中，使用均方误差定义目标函数；4)计算参数W关于损失函数的梯度：5)计算出上面的梯度后，使用SGD算法来更新参数，从而得到最优的Q值，即获得每次交互选取使得Q值最大的动作a。本发明方法利用预定义的方块统计特征改进了DQN网络训练Agent玩俄罗斯方块的性能。

摘要： 本发明公开了一种基于Siamese Network的茶砖识别方法及系统，属于孪生网络识别技术领域，本发明要解决的技术问题为如何利用茶砖的自身特点并结合孪生网络实现茶砖的防伪识别，采用的技术方案为：具体如下：S1、获取原始特征提取图和待识别特征提取图：在每一个茶砖生产形成之初对茶砖的特征纹路进行特征提取与分析得到五个局部原始特征提取图；当消费者买到新的茶砖时，获取茶砖的照片并提取相同位置的五个局部待识别特征图；S2、将得到的五个局部原始特征提取图作为该茶砖自身的防伪标识与该茶砖的参数信息一并储存到产品质量信息库中；S3、通过孪生网络比对五个局部原始特征提取图与五个局部待识别特征图之间的距离进行茶砖防伪标识识别并作出判定。

摘要： 本发明公开了一种用于加速Deep Q‑Network算法的硬件架构及其设计空间探索方法。硬件架构包括：通用处理器模块负责与外部环境进行交互和实现奖励函数的计算，也负责Deep Q‑Network算法经验池的维护；外部DDR存储器负责Deep Q‑Network算法的经验池的存储；AXI总线接口为通用AXI总线接口结构，负责实现通用处理器与FPGA可编程逻辑模块之间控制信号和数据信号的传递与反馈；Target Q模块负责实现Target Q网络的前向推理计算；Current Q模块负责实现Current Q网络的前向推理和反向传播。本发明在高度优化FPGA硬件架构下，实现Deep Q‑Network算法的实时计算。

摘要： 本发明涉及在碳支持铜网上生长Ag/SDS层：一种高效、可移动的荧光增强平台，具体公开了一种以铜网为基底的Ag/SDS nested‑network荧光增强材料及其制备方法，与现有技术相比，本发明在室温条件下通过简单的溶液浸入法制备出负载在铜网上的Ag/SDS nested‑network结构，得到的树枝状的银纳米薄层有望具有超高的荧光增强效应，且此类材料可随意移动和安置，有望实现工业化应用。

摘要： 本发明提供一种基于Deep Q‑network算法的多资源云作业调度方法，包括四个步骤：通过云环境采集资源的当前配置信息和作业的需求信息；所述资源的当前配置信息和作业的需求信息分别以矩阵图像表示，所述包括单元格，同一种彩色的单元格表示同一个作业，同一种彩色的单元格形成的矩形包括M×N个单元格，M表示资源数，N表示时间步；根据所述矩阵图像，利用深度学习方法得到高层语义信息；根据所述高层语义信息，利用强化学习方法完成资源的实时调度规划。本发明矩阵图像能完全清晰地表示资源与作业的状态。本方法还使用了深度强化学习，将深度与强化学习进行了结合，其中深度学习主要完成环境状态感知，而强化学习完成决策，实现状态到动作的映射。

摘要： 本发明属于生产计划技术领域，具体涉及一种基于Deep Q‑network深度强化学习的单件作业车间调度方法，包括以下步骤：(1)采用析取图的方法对作业车间调度环境进行建模，将调度决策问题转换为序贯决策问题，建立马尔科夫五元组模型，使用深度强化学习对该模型进行求解；(2)从作业车间调度的析取图环境中提取当前的状态；(3)采用卷积神经网络对动作值函数和目标值函数进行拟合；(4)采用18种启发式调度规则，作为强化学习的代理动作；(5)设计奖励函数对整个调度决策进行评估，使用DQN算法更新动作值函数的权重参数；(6)进行状态转移；(7)目标值函数的网络参数更新；本发明可以快速处理作业车间的调度问题，具有实时性强和灵活性高的优点。

摘要： 本发明提供一种基于Deep Q‑network算法的多资源云作业调度方法，包括四个步骤：通过云环境采集资源的当前配置信息和作业的需求信息；所述资源的当前配置信息和作业的需求信息分别以矩阵图像表示，所述包括单元格，同一种彩色的单元格表示同一个作业，同一种彩色的单元格形成的矩形包括M×N个单元格，M表示资源数，N表示时间步；根据所述矩阵图像，利用深度学习方法得到高层语义信息；根据所述高层语义信息，利用强化学习方法完成资源的实时调度规划。本发明矩阵图像能完全清晰地表示资源与作业的状态。本方法还使用了深度强化学习，将深度与强化学习进行了结合，其中深度学习主要完成环境状态感知，而强化学习完成决策，实现状态到动作的映射。