mseq

This function takes a training-by-training matrix of distance scores, and the testing-by-training matrix of distance scores, and augments them by adding ghost points to the minority class.

说明：  SLIC method is used to segment polarimetric SAR image and calculate classification accuracy

用Hausdorff距离对两角点集进行配准，得到点集间的仿射变换，从而实现图像的自动配准。此算法以角点作为Hausdorff距离的配准特征，与直接选用边缘来配准的方法相比较，大大减小计算量。(Hausdorff distance on the corners with a point set registration, be affine transformation between sets in order to achieve automatic image registration. This algorithm is based corner as the Hausdorff distance matching characteristics, and use Edge to direct registration method of comparison, greatly reduced computation.)

通过恢复低秩矩阵得到显著性图，其中详细的运用了mean-shift算法，可以很好的了解mean-shift算法的使用，同时还包含了RPCA算法。(Significantly through the recovery of low rank matrix diagram, which detailed the use of mean-shift algorithm, can be a very good understanding of the use of the mean-shift algorithm, also contains RPCA algorithm.)

Hamming 15, 11 Viterbi encoding and Decoding

随机游走类似布朗运动，就是随机的向各个方向走吧。 虽然代码没什么技术含量，不过产生的图像实在太漂亮了，所以还是贴上来吧。 产生的图像：(Similar to Brownian motion random walk, is random and go in all directions. Although the code is no technical content, but the resulting image is too beautiful, so it is still Tieshanglai it. graphics layer.)

这个函数也要被melcepst函数调用，用来进行计算Mel倒谱系数。这个函数的作用是构造mel滤波器组。(This function must be called melcepst function is used to calculate Mel cepstral coefficients. The role of this function is to construct mel filter bank.)

NP是美国匹兹堡大学的T.L.Saaty 教授于1996年提出了一种适应非独立的递阶层次结构的决策方法，它是在网络分析法（AHP）基础上发展而形成的一种新的实用决策方法。其关键步骤有以下几个： 1 确定因素，并建立网络层和控制层模型。 2 创建比较矩阵。 3 按照指标类型针对每列进行规范化。 4 求出每个比较矩阵的最大特征值和对应的特征向量。 5 一致性检验。如果不满足，则调整相应的比较矩阵中的元素。 6 将各个特征向量单位化(归一化)，组成判断矩阵。 7 将控制层的判断矩阵和网络层的判断矩阵相乘，得到加权超矩阵。 8 将加权超矩阵单位化(归一化)，求其K次幂收敛时的矩阵。其中第j列就是网络层中各元素对于元素j的极限排序向量。 (NP is a professor at the University of Pittsburgh TLSaaty presented in 1996, an adaptation of non-independent Hierarchy of decision-making method, which is the analytic network process (AHP) formed on the basis of the development of a new and practical decision-making method . The key steps are the following: A determining factor, and a network layer and control layer model. 2 create a comparison matrix. For each of the three types of indicators in accordance with normalized columns. 4 find the maximum for each comparison matrix eigenvalue and the corresponding eigenvectors. 5 consistency test. If not satisfied, then the comparison to adjust the corresponding matrix elements. 6 will each feature vector units of (normalized), to determine the composition of matrix. 7 to determine the control layer and network layer to determine matrix matrix multiplication, to be weighted super-matrix. 8 of the weighted super-matrix units of (normalized), seeking the powe)

龙哥库塔法的matlab实现方法，简单易用(Long Ge Kuta method matlab implementation, easy to use)

用变形核密度模型对数据的统计分布进行估计，常用于金融信用分析、财产保险精算等领域，是一个很有效、常用的非参数模型(Deformation of the nuclear density model of the statistical distribution of the data to estimate the financial credit analysis, property insurance and actuarial fields, is a very effective and commonly used non-parametric model)