基于词-标签概率的多标签文本分类研究

针对多标签文本分类任务中如何有效地提取文本特征和获取标签之间潜在的相关性问题,提出一种CNN(convolutional neural networks)结合Bi-LSTM (bi-directional long short-term memory)的模型.首先,通过CNN网络和最大池化提取文本的特征;然后,利用训练的Labeled-LDA(labeled latent dirichlet allocation)模型获取所有词与标签之间的词-标签概率信息;接着,使用Bi-LSTM网络和CNN网络提取当前预测文本中每个词的词-标签信息特征;最后,结合提取的文本特征,预测与当前文本相关联的标签集.实验结果表明,使用词-标签概率获取文本中词与标签之间的相关性信息,能够有效提升模型的F1值.     

IRS辅助的MISO-SWIPT安全速率最大化算法

为增强系统的稳定性,基于智能反射面（intelligent reflection surface,IRS）辅助安全通信,为用户采用功率分配架构的无线携能传输（simultaneous wireless information and power transfer,SWIPT）技术以提高用户安全速率并降低用户的能量消耗,建立了系统安全速率最大化的目标函数模型;考虑IRS相移约束、基站最大发射功率、功率分配比率约束和用户最低能量采集约束,提出了一种基于交替优化的安全速率最大化算法。利用连续凸近似算法,将目标函数转化为凸的形式;采用交替优化,对耦合变量进行解耦处理;提出一种波束赋形算法以实现安全速率最大化。仿真表明,提出的算法能够有效提升IRS辅助的SWIPT系统的安全速率,至少比其他算法提高了2.63bps/Hz。     

IRS辅助WPCN用户公平性与系统吞吐量优化方案

为了解决无线供电网（wireless powered communication network,WPCN）中用户公平性不足与系统灵活性差的问题,提出一种基于智能反射面（intelligent reflecting surface,IRS）的优化方案。该方案直接将公平指数（fairness index,FI）和系统吞吐量加权相加作为优化目标,建立了用户公平性与系统吞吐量的联合优化问题。由于所建立的优化问题是非凸问题,因此,将优化问题拆分为时隙长度和IRS反射系数优化子问题,然后交替求解。针对时隙长度优化子问题,使用逐次凸逼近（successive convex approximation,SCA）算法求解;针对IRS反射系数优化子问题,使用闭式解、SCA和半定松弛（semi definite relaxation,SDR）算法求解。仿真结果表明,所提方案在用户公平性和系统灵活性方面优于其他方案。     

聚氨酯／聚甲基丙烯酸甲酯互穿聚合物网络的研究

本文对(蓖麻油-聚乙二醇)聚氨酯/聚甲基丙烯酸甲酯(PU/PMMA)互穿聚合物网络(IPN)体系进行了研究.实验结果表明:IPN的力学性能受网络化学构型控制.在PU/PMMA为25/75时,应力-应变曲线出现屈服点,产生类似皮革行为,并在此点抗张强度最高,表现明显的协同效应.动态力学性能及电镜均证明该体系相分离十分严重,通过加入扩链剂及网间交联剂能显著改善两组分相容性,其中网间交联剂效果最好。     

Event-triggered sliding mode control for switched genetic regulatory networks with persistent dwell time

This paper focus on the event-triggered sliding mode controller design for discrete-time switched genetic regulatory networks (GRNs) with persistent dwell time (PDT) switching. Firstly, the observation error dynamics of switched GRNs with PDT is constructed in the light of event-triggered sliding mode control (SMC) scheme. Next, sufficient conditions are derived to ensure the exponential stability of the augmented plant. Moreover, an event-triggered SMC law is synthesized to impel the system trajectories onto the sliding surface in a finite time. Finally, a verification example is provided to illustrate the effectiveness and potential of the proposed method.     

离散时间细胞神经网络中的分岔和混沌

研究了离散时间细胞神经网络（ＤＴＣＮＮ）中的分岔和混沌,ＤＴＣＮＮ中的细胞与连续时间ＣＮＮ中的细胞相似,具有局部耦合,其输出方程是罗切斯蒂方程。分别讨论了２种结构（有边界和无边界）的ＤＴＣＮＮ的混沌特性,ＤＴＣＮＮ的混沌与全局耦合系统一的混沌相似,但由于其局部耦合的结构,其混沌与后者有所不同。提出了有关分岔与混沌的一些有兴趣的理论问题和实际问题．     

Retrieval and chaos in extremely dilutedQ-Ising neural networks

Using a probabilistic approach, the deterministic and the stochastic parallel dynamics of aQ-Ising neural network are studied at finiteQ and in the limitQ. Exact evolution equations are presented for the first time-step. These formulas constitute recursion relations for the parallel dynamics of the extremely diluted asymmetric versions of these networks. An explicit analysis of the retrieval properties is carried out in terms of the gain parameter, the loading capacity, and the temperature. The results for theQ network are compared with those for theQ=3 andQ=4 models. Possible chaotic microscopic behavior is studied using the time evolution of the distance between two network configurations. For arbitrary finiteQ the retrieval regime is always chaotic. In the limitQ the network exhibits a dynamical transition toward chaos.     

1/d corrections for interacting spinless fermions: One-particle properties

One-particle properties of the spinless fermion model with repulsion at half filling are calculated within an approach correct to first order in the inverse of the lattice dimensiond. Continuity of the limitd requires a scaling of the nearest-neighbour hopping proportional to and of the nearest-neighbour interaction proportional to 1/d. Due to this scaling the Hartree approximation becomes exact in infinite dimensions. We show that 1/d corrections comprise the Fock diagram and the local correlation diagram in the self-consistent Dyson equation. This approach is applied to simple-cubic systems in dimensiond=1, 2 and 3. Ground state properties and the charge-density wave phase diagram are calculated. AtT=0 the inclusion of 1/d terms gives only small corrections to the leading Hartree contribution ind=2, 3. ForT>0, however, the 1/d corrections are important. They lead to a non-negligible reduction of the critical temperature. Ind=1 the 1/d corrections are very large, but they do not succeed in removing the spurious phase transition atT>0. The 1/d approach provides a good and tractable approximation ind=3 and probably ind=2, which allows also further systematic improvement.     

采用HK模型的防空反导作战网络建模分析

采用HK网络模型对防空反导作战网络进行建模,对网络度分布、集群系数、平均距离的统计特性进行分析。将网络模型与典型树状结构的防空反导网络和基于BA无标度模型的网络进行比较,表明基于HK模型的防空反导作战网络具有较好的集群系数和无标度特性,同时表明改变TF连接概率进行防空反导网络设计具有更好的灵活性。     

无线多媒体传感器网络自适应拥塞控制算法

针对无线多媒体传感器网络(WMSN)中多对一通信时产生的网络拥塞问题,提出了一种自适应的WSMN网络拥塞控制算法ACCP。通过结合速率控制和资源调度,并采用分簇的网络结构,根据簇首及簇内的拥塞指标,来分别启动对应的拥塞控制机制:当簇首发生短时间拥塞时,就启动属于资源调度的网络内存储管理机制,来暂时减缓网络内过多数据包;但当簇中的存储节点也无法容纳过量的数据包时,速率控制就启动,让流量减缓下来,并且只调整数据实时性要求较低的数据流的速率,以达到控制流量、减缓甚至消除网络拥塞的目的。仿真结果显示:ACCP在传送速率不同下,可以比InS、HCCP更有效的控制网络拥塞的情况,而在比较缓存容量不同的情况下,虽然ACCP只比HCCP能够稍微改善网络拥塞的情况,但却能够大幅度改善InS的数据包丢失率。