基于组合预测法的共享单车租赁需求量预测

共享单车的租赁需求量预测对于单车企业提升运营效率十分必要,是单车再调度的前提。为了更加准确地预测出共享单车的租赁需求量,本文结合随机森林、XGBoost、GBDT三类数据驱动预测算法的优点,提出了一种基于向量投影法的加权对数平均组合模型。定义了组合模型的优性,非劣性,劣性的概念。并证明了该方法至少是一种非劣性的预测方法。通过将该方法运用于现实问题中,以解决实际单车租赁需求量预测问题。实例研究发现:该方法在单车租赁需求量预测中可以为优性预测模型, 能够对单车再调度起到正向作用。该方法可以为单车租赁需求量预测的相关研究提供一种切实有效的解决方向。     

一维随机介质内光波模式的频率微分特性研究

本文用数值方法在有限尺寸的随机介质中求解麦克斯韦方程组,对一维随机激光器的频率微分特性进行了研究.结果表明每个随机散射颗粒都对介质内光波模式的频率形成有不同的贡献,散射颗粒尺寸和位置的微小变化会使介质内光波频率发生变化.处于局域化中心位置的介质膜,它们厚度的变化对输出频率的影响较大,局域化区域以外位置处的薄膜,它们的厚度的变化对频率的影响则非常微小.     

On a backward problem for fractional diffusion equation with Riemann-Liouville derivative

In the present paper, we study the initial inverse problem (backward problem) for a two-dimensional fractional differential equation with Riemann-Liouville derivative. Our model is considered in the random noise of the given data. We show that our problem is not well-posed in the sense of Hadamard. A truncated method is used to construct an approximate function for the solution (called the regularized solution). Furthermore, the error estimate of the regularized solution in L² and H^τ norms is considered and illustrated by numerical example.     

随机介质背散射二维Mueller矩阵的数值模拟与分析

推导了随机介质背散射Mueller矩阵的直接计算公式,并运用矢量Monte Carlo方法进行了数值模拟.结果表明随机介质背散射二维Mueller矩阵方位关系随散射系数的减小而增强,而与微粒大小关系不大;Mueller矩阵元素绝对值的空间分布随径向呈近似指数规律衰减,矩阵元素的方位变化具有周期性.对称系统的二维Mueller矩阵的花样图中仅有7幅独立,其余9幅可通过对称、旋转变换得到.     

基于混沌激光产生1 Gbit/s的随机数

利用光反馈半导体激光器产生的混沌激光作为随机数发生器的物理熵源,通过8位 ADC将熵源信息转化为二进制码,并经后续差分运算处理改善其随机性,最终获得了1 Gbit/s的随机数.所产生的随机数通过了NIST Special Publication 800-22的全部测试项. 关键词： 混沌激光 随机数发生器 半导体激光器 模数转换     

On the imaginary parts of chromatic roots

While much attention has been directed to the maximum modulus and maximum real part of chromatic roots of graphs of order $n$ (ie, with $n$ vertices), relatively little is known about the maximum imaginary part of such graphs. We prove that the maximum imaginary part can grow linearly in the order of the graph. We also show that for any fixed $\text{◂+▸}p\in \left(0,1\right)$, almost every random graph $G$ in the Erdös-Rényi model has a nonreal root.     

Resonance response of a single-degree-of-freedom nonlinear vibro-impact system to a narrow-band random parametric excitation

The resonant response of a single-degree-of-freedom nonlinear vibro-impact oscillator with a one-sided barrier to a narrow-band random parametric excitation is investigated. The narrow-band random excitation used here is a bounded random noise. The analysis is based on a special Zhuravlev transformation, which reduces the system to one without impacts, thereby permitting the applications of random averaging over "fast" variables. The averaged equations are solved exactly and an algebraic equation of the amplitude of the response is obtained for the case without random disorder. The methods of linearization and moment are used to obtain the formula of the mean-square amplitude approximately for the case with random disorder. The effects of damping, detuning, restitution factor, nonlinear intensity, frequency and magnitude of random excitations are analysed. The theoretical analyses are verified by numerical results. Theoretical analyses and numerical simulations show that the peak response amplitudes will reduce at large damping or large nonlinear intensity and will increase with large amplitude or frequency of the random excitations. The phenomenon of stochastic jump is observed, that is, the steady-state response of the system will jump from a trivial solution to a large non-trivial one when the amplitude of the random excitation exceeds some threshold value, or will jump from a large non-trivial solution to a trivial one when the intensity of the random disorder of the random excitation exceeds some threshold value.     

Random walks generated by affine mappings

This article is concerned with Markov chains on ^m constructed by randomly choosing an affine map at each stage, and then making the transition from the current point to its image under this map. The distribution of the random affine map can depend on the current point (i.e., state of the chain). Sufficient conditions are given under which this chain is ergodic.     

Stochastic flows on a countable set: Convergence in distribution

Consider a sequenceF ₁,F ₂,... of i.i.d. random transformations from a countable setV toV. Such a sequence describes a discrete-time stochastic flow onV, in which the position at timen of a particle that started at sitex isM _n(x), whereM _n =F _n F _n–1 F ₁. We give conditions on the law ofF ₁ for the sequence (M _n) to be tight, and describe the possible limiting law. an example called the block charge model is introduced. The results can be formulated as a statement about the convergence in distribution of products of infinite-dimensional random stochastic matrices. In practical terms, they describe the possible equilibria for random motions of systems of particles on a countable set, without births or deaths, where each site may be occupied by any number of particles, and all particles at a particular site move together.