Bayes分类器应用于IRIS数据集的例子

本人的模式识别课程VC源程序-my pattern recognition programs VC source

SOBI算法，采用二阶统计量进行盲分离。-SOBI algorithm, using second-order statistic for blind separation.

包括883分酒问题，图的遍历等人工智能基本算法-Including 883 hours of liquor problem, artificial intelligence, such as graph traversal algorithm

Using Simulated annealing to solve TSP

CPN算法的matlab实现，CPN人工神经网络-Matlab algorithm of network，The artificial neural network, network

基于Matlab的简单遗传算法和简单bp网络的程序实现-based on Matlab simple genetic algorithms and simple bp Network Program

he algorithm is equivalent to Infomax by Bell and Sejnowski 1995 [1] using a maximum likelihood formulation. No noise is assumed and the number of observations must equal the number of sources. The BFGS method [2] is used for optimization. The number of independent components are calculated using Bayes Information Criterion [3] (BIC), with PCA for dimension reduction.

遗传算法-理论和应用与软件实现，这是一本比较好的书-Genetic algorithms- theory and application and software implementation, this is a good book

遗传算法的源程序例子，对于初学者有很大帮助，建议多-Genetic algorithm source code examples are very helpful for beginners, the proposed multi-

提出了基于最小化估计的泛化误差界非优化估计方法。目前大多数的核参数选择方法都是通过极小化了来得到最优参数值，但是求解优化问题的计算代价相当的大，并且不能很好地体现数据的分布特征。本文采用非优化技术，通过极小化泛化误差来优化核及相关参数，由于直接计算最小半径和最大间隔，避免了对优化问题的直接求解，因此可以很好地降低计算代价。并且该方法直接从样本出发，可以很好地体现数据的分布特征，不管数据分布是否均匀都可以适用。给出了基于凸包估计的SVM核选择的模型及实现算法。 -Proposed based on minimizing the estimated generalization error bound of the non-optimal estimation method. Most of the nuclear parameter selection methods are to get the best by minimizing the parameter values, but the computational cost for solving optimization problems is quite large, and can not properly reflect the distribution characteristics of the data. In this paper, the non-optimization technology, by minimizing the generalization error to optimize the nuclear and related parameters, due to the direct calculation of the minimum radius and maximum interval, avoiding the direct solution of the optimization problem, it can very well reduce the computation cost. And that the method directly from the sample one can well reflect the distribution chara