COMSOL 5 for Engineers-Mehrzad Tabatabaian

支持向量机和BP神经网络都可以用来做非线性回归拟合，但它们的原理是不相同的，支持向量机基于结构风险最小化理论，普遍认为其泛化能力要比神经网络的强。大量仿真证实，支持向量机的泛化能力强于BP网络，而且能避免神经网络的固有缺陷——训练结果不稳定。本源码可以用于线性回归、非线性回归、非线性函数拟合、数据建模、预测、分类等多种应用场合(Support vector machines and BP neural network can be used for non-linear regression fit, but they is not the same principle, support vector machine based on structural risk minimization theory, generally considered its generalization ability than the neural network of strong . Inherent defect- a large number of simulation confirmed that generalization ability of support vector machine is stronger than the BP network, but also to avoid neural network training results unstable. The source code can be used for linear regression, nonlinear regression, nonlinear function fitting a variety of applications, data modeling, forecasting, classification)

说明：  利用GA遗传算法解决欺骗函数最优问题，具体问题描述如下，如有问题请与我联系(The deceptive functions are a family of functions in which there exists low-order building blocks that do not combine to form the higher-order building blocks. Here, a deceptive problem that consists of 25 copies of the order-4 fully deceptive function DF2 is constructed for this paper. DF2 can be described as follows: f(0000)=28 f(0001)=26 f(0010)=24 f(0011)=18 f(0100)=22 f(0101)=6 f(0110)=14 f(0111)=0 f(1000)=20 f(1001)=12 f(1010)=10 f(1011)=2 f(1100)=8 f(1101)=4 f(1110)=6 f(1111)=30 This problem has a maximal function value of 750.)

FCT_VOF AND YOUNGS_PLIC fully functioning source codes Debugged and fully functioning source codes Both can be used to calculate by choosing the icase for (1) the Zalesak rotational flow problem (2) the shear flow problem This complete set includes both FCT_VOF and Youngs_plic source code in f90. Including a input data file. The codes have been debugged and tested using MS-FORTRAN. Results can be obtained directly from the output files and loaded to get results. (FCT_VOF AND YOUNGS_PLIC fully functioning source codes Debugged and fully functioning source codes Both can be used to calculate by choosing the icase for (1) the Zalesak rotational flow problem (2) the shear flow problem This complete set includes both FCT_VOF and Youngs_plic source code in f90. Including a input data file. The codes have been debugged and tested using MS-FORTRAN. Results can be obtained directly from the output files and loaded to get results. Ps: statement CALL WRITER needed to be removed and replaced with usual output statement fsuch as: Write(*,*) x(I,j),y(I,j),fi(I,j)…. Etc., for user’s requirement. )

Python for Data Analysis DATA WRANGLING WITH PANDAS, NUMPY, AND IPYTHON

用蒙特卡洛模拟的方法计算美式期权的价格以及基本描述,实现了图像的加水印，去噪，加噪声等功能，包括最小二乘法、SVM、神经网络、1_k近邻法。( Monte Carlo simulation method of calculating the American option price and basic description, Realize image watermarking, de-noising, plus noise and other functions, Including the least squares method, the SVM, neural networks, 1 _k neighbor method.)

说明：  本书详细介绍了格子波尔兹曼法的原理，边界条件，具体模型及数值方法，非常适合学习与应用。(This book introduce the theory of lattice boltzmann methods alone with it's boundary conditions,models and numerical methods. It's really a book for us to learn and use.)

这是fotran的非线性有限元程序，包括1D，2D程序(fortan NFEM )

小波分解与最小二乘支持向量机，用于天线故障诊断。(Wavelet decomposition and least squares support vector machine for antenna diagnostics.)

用于研究粒子在晶体材料内输运性质，计算扩散系数的动力学蒙特卡洛程序( Kinetic Monte Carlo program is used to study particle transport properties in crystalline materials, the calculation of the diffusion coefficient Kinetic Monte Carlo program is used to study particle transport properties in crystalline materials, the calculation of the diffusion coefficient)

基于随机共振的微弱信号检测仿真，且包含求解郎之万方程的四阶龙格库塔算法源程序。(Fourth-order Runge-Kutta algorithm source simulation based on stochastic resonance in weak signal detection, and includes solving the Langevin equation.)