能量收集传感器网络中改进的能耗管理方案研究

能量收集传感器平台为网络协议的设计开辟了新的领域。为了支持网络的运行,能耗率不得高于能量收集率,否则,传感器节点会最终耗尽它们的能量。与资源处于静态的传统网络资源分配问题相比,充电速率的时间可变性带来了新的挑战。在本文中,我们首先研究了一种基于高效对偶分解和次梯度策略的算法QuickFix,计算出数据采样率和路由。然而,再次充电发生波动时的时标可能会快于传统方法的收敛时间,进而导致电池断电和溢出,这会造成采样丢失和能量收集机会丢失。为了解决这一动态问题,我们提出一种本地算法SnapIt,通过对采集率进行调节以维持电池电量在目标水平上。基于TOSSIM模拟器的性能评估表明,QuickFix和SnapIt联合起来可以跟踪网络瞬时最优效用,同时维持电池电量处于目标水平。与基于余压的IFRC相比,本文方法使总体数据速率平均提升42%,同时显著提升了网络效用。     

基于遗传神经网络的VAV空调系统预测模型

针对单纯的机理建模方法难以准确预测变风量空调系统（VAV）的参数,利用BP神经网络构建了变风量空调系统的预测模型,并将遗传算法与BP网络相结合,提出运用遗传算法对神经网络的权值和阈值进行遗传搜索,寻优后再进行BP运算,以克服BP算法收敛速度慢、易陷入局部解的缺点。通过实验平台采集了大量数据对所建模型进行训练和验证,结果表明,模型对空调送风参数以及房间温湿度的预测结果与实测数据能很好拟合,精确度高,泛化能力强。     

带反应扩散和分布时滞的神经网络的稳定性

文章中我们研究了带反应扩散和分布时滞的神经网络,利用M矩阵的性质和不等式技巧,给出了系统平衡点一致稳定的判别准则.     

Parabolic scaling of tree-shaped constructal network

We investigate the multi-scale structure of a tree network obtained by constructal theory and we propose a new geometrical framework to quantify deviations from scale invariance observed in many fields of physics and life sciences. We compare a constructally deduced fluid distribution network and one based on an assumed fractal algorithm. We show that: (i) the fractal network offers lower performance than the constructal object, and (ii) the constructal object exhibits a parabolic scaling explained in the context of the entropic skins geometry based on a scale diffusion equation in the scale space. Constructal optimization is equivalent to an equipartition of scale entropy production over scale space in the context of entropic skins theory. The association of constructal theory with entropic skins theory promises a deterministic theory to explain and build optimal arborescent structures.     

Parameter redundancy in neural networks: an application of Chebyshev polynomials

This paper deals with feedforward neural networks containing a single hidden layer and with sigmoid/logistic activation function. Training such a network is equivalent to implementing nonlinear regression using a flexible functional form, but the functional form in question is not easy to deal with. The Chebyshev polynomials are suggested as a way forward, providing an approximation to the network which is superior to Taylor series expansions. Application of these approximations suggests that the network is liable to a ‘naturally occurring’ parameter redundancy, which has implications for the training process as well as certain statistical implications. On the other hand, parameter redundancy does not appear to damage the fundamental property of universal approximation.      

Development of projection and artificial compressibility methodologies using the approximate factorization technique

Predictions for two-dimensional, steady, incompressible flows under both laminar and turbulent conditions are presented. The standard k-? turbulence model is used for the turbulent flows. The computational method is based on the approximate factorization technique. The coupled approach is used to link the equations of motion and the turbulence model equations. Mass conservation is enforced by either the pseudocompressibility method or the pressure correction method. Comparison of the two methods shows a superiority of the pressure correction method. Second- and fourth-order artifical dissipation terms are used in order to achieve good convergence and to handle the turbulence model equations efficiently. Several internal and external test cases are investigated, including attached and separated flows.     

神经网络中的退火：非时齐情形

本文考虑神经网络经过非时齐退火后的极限行为,我们给出了神经网络经过非时齐退火后的分布,证明了经过非时齐退火后网络在组态空间的一个子集上是遍历的。从而网络经非时齐退火后并不能以概率为1地恢复起一个记忆,进一步地,我们证明了既使只贮存了一个最强的记忆η,网络经非时齐退火仍不能以概率为1地恢复它。     

基于神经网络的波达方向实时估计问题探讨

本文在无约束非线性规划神经网络模型的基础上,提出了一种实时估计空间多个信号到达方向即波达方向的方法,并给出了实例验证。     

可作废推理

推理的不确定性来自信息或知识的不完全,这种不确定性几乎是无处不存在的,它使得修改乃至废除推理过程中的结论显得十分重要,从而也使可作废推理成为智能行为(不论是人或机器的)的显著特征。因此,研究推理的不确定性并使之形式化成为人工智能中最新、最热也最难的方向。本文扼要介绍可作废推理的研究途径、发展、特征和有待解决的问题;同时简介作者的一些新近研究结果,以期引起同行的更深入、更有价值的探索。     

64路智能数字转接机的软件设计

智能数字转接机（ＩＤＳ）是ＴＥＬＩＮＥＴ（ＴＥｌｅｐｈｏｎｅＬＩｎｅＮＥＴｗｏｒｏｋ）网络组成中的核心部件。通过它可以完成计算机基带数据通信双方电话线物理链路的智能转接。笔者继１６路４通道ＩＤＳ研制成功并相继应用在几个单位的ＴＥＬＩＮＥＴ网络中后，又开发出了６４路８通道ＩＤＳ，阐述了它的软件设计和实现