Origin-Destination Demand Reconstruction Using Observed Travel Time under Congested Network

Networks and Spatial Economics - Two bi-level models to reconstruct origin-destination (O-D) demand under congested network are explored in terms of the observed link and route travel times, where...     

The global system-ranking efficiency model and calculating examples with consideration of the nonhomogeneity of decision-making units

Data envelopment analyses have been widely used to evaluate the relative efficiency of decision-making units (DMUs). However, the traditional data envelopment analysis model has not considered the problem of DMUs' nonhomogeneity. If nonhomogeneous DMUs are evaluated under the same production frontier, conclusions may not be precise. For example, some DMUs' input redundancy and output deficit cannot be adjusted as per planning results, which may lead to mistakes in management. This paper loosens the assumption of DMU homogeneity and builds a global system-ranking efficiency model based on existing literature, which divides the problem of DMUs' nonhomogeneity into external nonhomogeneity and internal homogeneity. Data have been collected from 114 listed enterprises in China's solar power industry, and the analysis results indicate that this paper's model is stable and reliable and can be used as a reference for production managers.     

非视距环境下基于RSS-TOA的定位算法

在无线传感器网络定位系统中,尤其是在室内定位中,非视距（NLOS）误差的存在使定位性能急剧下降。为克服非视距传播带来的定位误差,提出了一种针对非视距环境下联合接收信号强度（RSS）和到达时间（TOA）的定位算法。该方法首先通过 RSS和 TOA的测量结果建立关于目标位置的非凸优化问题,然后通过二阶锥松弛理论,将原始的非凸优化问题转换为一种凸优化问题,由此能够快速得到原问题的一个次优解。通过计算机模拟仿真验证,新方法的估计精度更高,性能更好。     

基于稀疏编码局部时空描述子的动作识别方法

针对已有动作识别算法训练速度慢且识别精度不高等问题,提出了基于稀疏编码局部时空描述子的动作识别方法。该方法首先对深度图像进行法线提取,同时应用基于运动能量的自适应时空金字塔对动作帧分块;然后局部聚集法线,得到显著性局部时空描述子;对局部时空描述子进行稀疏编码得到一组字典向量来重构样本数据;最后利用简化粒子群（sPSO）优化SVM分类器找到最适合样本数据的分类模型。实验在MSRAction3D和MSRGesture3D公开数据集上达到了93.80%和95.83%的识别率,且训练速度较传统方法有明显提升,证明了该方法的有效性和鲁棒性。     

制造物联环境下智能仓库货位分配模型

仓库货位合理规划与分配是减少产品存取时间、提高仓库作业效率的关键。在构建基于制造物联技术的智能仓库环境下,针对多品种智能仓库的货位分配问题,建立了考虑多规则约束的多目标智能仓库货位分配模型,提出一种模型求解的改进遗传算法。实验表明：模型和算法能找到有效的仓库货位分配方案,验证了其有效性。     

基于拉普拉斯中心性和密度峰值的无参数聚类算法

针对聚类算法的聚类中心选取需要人工参与的问题,提出了一种基于拉普拉斯中心性和密度峰值的无参数聚类算法（ALPC）。首先,使用拉普拉斯中心性度量对象的中心性;然后,使用正态分布概率统计方法确定聚类中心对象;最后,依据对象到各个中心的距离将各个对象分配到相应聚类中心实现聚类。所提算法克服了算法需要凭借经验参数和人工选取聚类中心的缺点。在人工数据集和真实数据集上的实验结果表明,与经典的具有噪声的基于密度的聚类方法（DBSCAN）、密度峰值聚类（DPC）算法以及拉普拉斯中心峰聚类（LPC）算法相比,ALPC具有自动确定聚类中心、无参数的特点,且具有较高的聚类精度。     

A Fast Finite Difference Method for a Continuous Static Linear Bond-Based Peridynamics Model of Mechanics

The peridynamic nonlocal continuum model for solid mechanics is an integro-differential equation that does not involve spatial derivatives of the displacement field. Several numerical methods such as finite element method and collocation method have been developed and analyzed in many articles. However, there is no theory to give a finite difference method because the model does not involve spatial derivatives of the displacement field. Here, we consider a finite difference scheme to solve a continuous static bond-based peridynamics model of mechanics based on its equivalent partial integro-differential equations. Furthermore, we present a fast solution technique to accelerate Toeplitz matrix-vector multiplications arising from finite difference discretization respectively. This fast solution technique is based on a fast Fourier transform and depends on the special structure of coefficient matrices, and it helps to reduce the computational work from \(O(N^{3})\) required by traditional methods to O(Nlog\(^{2}N)\) and the memory requirement from \(O(N^{2})\) to O(N) without using any lossy compression, where N is the number of unknowns. Moreover, the applicability and accuracy of the scheme are demonstrated by numerical experiments to support our theoretical analysis.     

低秩稀疏图嵌入的半监督特征选择

目的 特征降维是机器学习领域的热点研究问题。现有的低秩稀疏保持投影方法忽略了原始数据空间和降维后的低维空间之间的信息损失,且现有的方法不能有效处理少量有标签数据和大量无标签数据的情况,针对这两个问题,提出基于低秩稀疏图嵌入的半监督特征选择方法（LRSE）。方法 LRSE方法包含两步：第1步是充分利用有标签数据和无标签数据分别学习其低秩稀疏表示,第2步是在目标函数中同时考虑数据降维前后的信息差异和降维过程中的结构信息保持,其中通过最小化信息损失函数使数据中有用的信息尽可能地保留下来,将包含数据全局结构和内部几何结构的低秩稀疏图嵌入在低维空间中使得原始数据空间中的结构信息保留下来,从而能选择出更有判别性的特征。结果 将本文方法在6个公共数据集上进行测试,对降维后的数据采用KNN分类验证本文方法的分类准确率,并与其他现有的降维算法进行实验对比,本文方法分类准确率均有所提高,在其中的5个数据集上本文方法都有最高的分类准确率,其分类准确率分别在Wine数据集上比次高算法鲁棒非监督特征选择算法（RUFS）高11.19%,在Breast数据集上比次高算法RUFS高0.57%,在Orlraws10P数据集上比次高算法多聚类特征选择算法（MCFS）高1%,在Coil20数据集上比次高算法MCFS高1.07%,在数据集Orl64上比次高算法MCFS高2.5%。结论 本文提出的基于低秩稀疏图嵌入的半监督特征选择算法使得降维后的数据能最大限度地保留原始数据包含的信息,且能有效处理少量有标签样本和大量无标签样本的情况。实验结果表明,本文方法比现有算法的分类效果更好,此外,由于本文方法基于所有的特征都在线性流形上的假设,所以本文方法只适用于线性流形上的数据。     

Sub-Femtojoule-Energy-Consumption Conformable Synaptic Transistors Based on Organic Single-Crystalline Nanoribbons

Inspired from powerful functionalities of human brain, artificial synapses are innovated continuously for the construction of brain-like neuromorphic electronics. The quest to rival the ultralow energy consumption of biological synapses has become highly compelling, but remains extremely difficult due to the lack of appropriate materials and device construction. In this study, organic single-crystalline nanoribbon active layer and elastic embedded photolithographic electrodes are first designed in synaptic transistors to reduce energy consumption of single device. The minimum energy consumption (0.29 fJ) of one synaptic event is far lower than that of biological synapse (10 fJ). Notably, sub-femtojoule-energy-consumption synaptic transistors can simulate various biological plastic behaviors even under different tensile and compressive strains, offering a new guidance for the construction of ultralow-energy-consuming neuromorphic electronic devices and the development of flexible artificial intelligence electronics in the future.     

Crumple Durable Ultraflexible Organic Solar Cells with an Excellent Power-per-Weight Performance

Ultraflexible and ultra-lightweight organic solar cells (OSCs) have attracted great attention in terms of power supply in wearable electronic systems. Here, ultrathin and ultra-lightweight OSCs, with a total thickness of less than 3 µm, with excellent mechanical properties in terms of their flexibility and ability to be stretched are demonstrated. A stabilized power conversion efficiency (PCE) of 15.5% and unprecedented power-per-weight of 32.07 W g⁻¹ at a weight of 4.83 g m⁻² is achieved, which represents one of the best-performing OSCs based on ultrathin foils substrate reported to date. The ternary strategy introduces the third component of amorphous conformation of the PC₇₁BM molecule, which can slightly reduce crystallization and aggregates without decreasing the electron mobility, thereby reducing rigidity and brittleness of the active layer. The increase in the ductility of the active layer significantly improves the mechanical flexibility of the device, resulting in over 90% retention in the PCE after 200 stretching–compression cycles. In addition, the ternary device exhibits excellent stability when stored in a N₂-filled glove box, resulting in the PCE retaining over 95% of its initial efficiency even after 1000 h. This ultraflexible and ultra-lightweight photovoltaic foils constitute a major step toward the integration of power supply into malleable electronic textiles.