psoToolbox

自适应中的牛顿法。结论，当u靠近0.5时，wk 迅速增加到最佳权， 当u小于0.5时，wk 沿着一个方向单向靠近最佳权；u大于0.5时wk沿着起始点指向最佳权的方向左右振荡并向最佳权收敛 ！ (The Newton method adaptive. Conclusion, when u close to 0.5, wk quickly right to the best, when u is less than 0.5, wk along the a direction close to the best the right way u greater than 0.5 wk point along the starting point about the best direction for the right to the right of oscillation and convergence of the best!)

鲁棒自适应matlab程序和报告，验证已通过(Robust Adaptive matlab procedures and reporting, verification has passed)

Book about Fiber Optic Systems and Measurements

cavity flow Matlab code.

Using Advanced Signal Processing Techniques on Ambient Vibration Records of a Structure

matlab 做的一个GUI，用于模拟电子琴(matlab to do a GUI for simulation flower)

一种基于TV 整体变分算法的图像修复的MATLAB代码(INPAINTING)

测试可以跑，根据自己情况修改下函数即可. NSGA-III 首先定义一组参考点。然后随机生成含有 N 个（原文献说最好与参考点个数相同）个体的初始种群，其中 N 是种群大小。接下来，算法进行迭代直至终止条件满足。在第 t 代，算法在当前种群 Pt的基础上，通过随机选择，模拟两点交叉（Simulated Binary Crossover，SBX）和多项式变异 产生子代种群 Qt。Pt和 Qt的大小均为 N。因此，两个种群 Pt和 Qt合并会形成种群大小为 2N 的新的种群 Rt=Pt∪Qt。 为了从种群 Rt中选择最好的 N 个解进入下一代，首先利用基于Pareto支配的非支配排序将 Rt分为若干不同的非支配层（F1，F2等等）。然后，算法构建一个新的种群St，构建方法是从 F1开始，逐次将各非支配层的解加入到 St，直至 St的大小等于 N，或首次大于 N。假设最后可以接受的非支配层是 L层，那么在 L+ 1 层以及之后的那些解就被丢弃掉了，且 St FL中的解已经确定被选择作为 Pt+1中的解。Pt+1中余下的个体需要从 FL中选取，选择的依据是要使种群在目标空间中具有理想的多样性。(The test can run and modify the function according to its own situation. NSGA-III first defines a set of reference points. Then the initial population containing N individuals (preferably the same number of reference points as the original literature) was randomly generated, where N was the size of the population. Next, the algorithm is iterated until the termination condition is satisfied. On the basis of current population Pt, the algorithm simulates two-point crossover (SBX) and polynomial mutation to produce offspring population Qt by random selection.)

对C++语言的反思，阐述c++语言的精髓，及基于作者多年编程经验总结的体会(think the core of c++ programming language)

快速计算一维信号的塔式谱峭度图，选择谱峭度最大值区间，方便确定所需滤波器中心频率及带宽。(Computes the fast kurtogram of signal x.This fast algorithm uses a pyramidal decomposition of the signal into a user-specified number of levels (a good choice is nlevel = 8). The first level is the classical kurtosis of the signal, the second level is the spectral kurtosis in 2 octaves, the third level the spectral kurtosis in 4 half-octaves, etc. Non-integer levels refer to divisions into third octaves. Here below is an illustration on how the frequency band is successively split when increasing the level)