指派问题

ADAM （Adaptive Moment Estimation）是另外一种自适应学习率算法，它结合动量梯度 下降法，在不同参数方向上采用不同学习率，保留前几次迭代的梯度，能够很好 的适应于稀疏数据。(ADAM (Adaptive Moment Estimation) is another adaptive learning rate algorithm, which combines momentum gradient. The descent method, which uses different learning rates in different parameter directions and retains the gradients of previous iterations, is very good. It is suitable for sparse data.)

Description 设a[0…n-1]是一个包含n个数的数组，若在i<j的情况下，有a[i]>a[j]，则称(i, j)为a数组的一个逆序对（inversion）。 比如 <2,3,8,6,1> 有5个逆序对。请采用类似“合并排序算法”的分治思路以O(nlogn)的效率来实现逆序对的统计。 一个n个元素序列的逆序对个数由三部分构成： （1）它的左半部分逆序对的个数，（2）加上右半部分逆序对的个数，（3）再加上左半部分元素大于右半部分元素的数量。 其中前两部分（1）和（2）由递归来实现。要保证算法最后效率O(nlogn)，第三部分（3）应该如何实现？ 此题请勿采用O(n^2)的简单枚举算法来实现。 并思考如下问题： （1）怎样的数组含有最多的逆序对？最多的又是多少个呢？ （2）插入排序的运行时间和数组中逆序对的个数有关系吗？什么关系？ 输入格式 第一行：n，表示接下来要输入n个元素，n不超过10000。 第二行：n个元素序列。 输出格式 逆序对的个数。 输入样例 5 2 3 8 6 1 输出样例 5(Set a[0... N-1] is a n array containing n numbers. If there is a [i] > a [j] i n the case of I < j, then (i, j) is a n inversion pair of a array. For example, <2,3,8,6,1> has five reverse pairs. Please use the idea of "merge sorting algorithm" to achieve the statistics of inverse pairs with O (nlogn) efficiency. The number of inverse pairs of a sequence of n elements consists of three parts: (1) The number of reverse pairs in the left half, (2) the number of reverse pairs in the right half, (3) the number of elements in the left half is greater than that in the right half. The first two parts (1) and (2) are implemented by recursion. To ensure the final efficiency of the algorithm O (nlogn), how should the third part (3) be implemented? Do not use O (n ^ 2) simple enumeration algorithm to solve this problem.)

数值分析中运用matlab用hermit插值多项式逼近函数(Numerical analysis using matlab function hermit interpolation polynomial approximation)

薄板样条，一种很好的拟合曲线的方法。是最近比较常用的拟合方法。(thin plate spline is so nice fitting method. Features: (1) support arbitrary control points (2) allow regularization (3) estimate the warp parameters from corresponding pairs )

利用克拉茂克郎尼变换，求得表面等离子体相位虚数部分再不同波长或入射角度的值(estimation of the imaginary part via Krammers-Kronig relarions)

为了进一步提高直接序列／跳频（ＤＳ／ＦＨ）扩频系统的抗干扰能力，基于小波包变换结合递归最小二乘算法设计 了一种变换域自适应干扰抑制算法，该算法采用小波包分解定位窄带干扰，递归最小二乘算法抑制窄带干扰。通过蒙特卡 罗仿真分析在增加抗干扰模块后，ＤＳ／ＦＨ系统工作在准静态时，在不同信噪比条件下抗窄带干扰性能。仿真结果表明：该 算法具有较强的自适应性以及抗窄带干扰能力，其性能优于传统的直接置零法，适用于多音干扰下的恶劣通信环境。(In order to further improve the anti-jamming capability of direct sequence/frequency-hopping spread spectrum system (DS/FH), based on wavelet packet transform combined with recursive least squares algorithm design a transform domain adaptive interference suppression algorithm, the algorithm uses wavelet packet decomposition positioning narrowband interference, narrowband interference suppression recursive least squares algorithm. By Monte Carlo simulation analysis to increase the anti-jamming module DS/FH system work in the quasi-static, in different SNR anti narrowband interference performance. Simulation results show that: the algorithm has strong adaptability and resistance to narrowband interference ability, its performance is superior to the traditional method of direct zero, poor communication environment suitable for multi-tone interference.)

随机过程基础课程，包含泊松过程、马尔可夫链、鞅等经典方法，清晰详细，有案例讲解(Stochastic Processes basic courses, including Poisson process, Markov chain, martingale and other classical methods, clear and detailed, there is case to explain)

矩阵转换问题 行与列的个数均为n 的矩阵称为n 阶方阵。假定矩阵中的每个元素的值在0到9之间，则可以将矩阵中的所有元素按行依次排列得到一个“单行矩阵字符串”。(Matrix conversion, the number of rows and columns of the matrix n are called n-order square. Assume that each element of matrix value in the 0 to 9, you can press all the elements of the matrix rows in order of priority for a " single matrix string." )

本程序用于生成高质量的网格，可用于有限元，有限体积等方法的前处理。(This procedure is used to generate high-quality mesh, can be used for finite element, finite volume methods such as pre-processing.)

电磁场与微波技术下的射线追踪技术。应用c语言编程(Electromagnetic Field and Microwave Technology under the ray tracing technique. Application c programming language)