kmeans算法,许多同学做毕设需要用到

这个程序是在windows平台下运行game of life最好的平台。它不仅对细胞空间没有限制、运行速度非常快、界面很友好，而且能够为你提供大量的初始细胞分布图案，你还可以任意组合这些图案达到你象要的构形，然后让这些构形动起来组合出更加神奇的生命游戏的世界。-the procedure is the windows platform running game of life the best platform. It not only on the cell space is no restriction, running very fast, very friendly interface, but you have to provide a lot of initial cell distribution patterns, you can design any combination of these to you as to the configuration, then let these configuration movement combined to make life more wonderful game world.

本人这学期写八数码作业，采用三种人工智能算法，在vc2003开发完成-I write this semester eight digital operations, in three artificial intelligence algorithm, developed in vc2003

支持向量机(SVM)的SMO算法实现。包括matlab源代码文件、支持向量机的pdf文档，源代码word文档。-Support Vector Machine (SVM) algorithm of SMO. Including the matlab source code files, Support Vector Machines pdf documents, source code for word documents.

基于人工神经网络的房地产项目评判模型，值得一看-based on artificial neural network evaluation model for real estate projects, an eye-catcher

由于K－均值聚类算法局部最优的特点，而模拟退火算法理论上具有全局最优的特点。因此，用模拟退火算法对聚类进行了改进。20组聚类仿真表明，平均每次对K结果值改进8次左右，效果显著。下一步工作：实际上在高温区随机生成邻域是个组合爆炸问题（见本人上载软件‘k－均值聚类算法’所述），高温跳出局部解的概率几乎为0，因此正考虑采用凸包约束进行模拟聚类，相关工作正在进行。很快将奉献给各位朋友。-as K-means clustering algorithm for optimal local characteristics, and simulated annealing algorithm theory with the characteristics of the global optimum. Thus, simulated annealing algorithm for clustering improvements. Cluster Group of 20 simulations show that the average value of K results improved about eight times, the results are obvious. The next step : In fact, in high temperature generated random neighborhood is a combination of explosives (see my software on the "k-means clustering algorithm" mentioned above), high-temperature solution of partial out almost zero probability, it is considering the use of convex hull bound for simulation cluster, the work under way . Soon dedication to the ladies.

是vc++编写的BP神经网络预测源代码，在预测的方向需要的通知下载吧-Is vc++ Prepared by BP neural network prediction source code, in the prediction of the direction you need to download the notice

演化程序――遗传算法和数据编码的结合 具体介绍了遗传算法的应用-evolutionary process-- genetic algorithms and data encoding the specific combination of the application of genetic algorithms

人工智能实验§贝叶斯网络的因果推理关系§ 熟悉掌握Bayes定理，学习贝叶斯网络的因果推理-artificial intelligence experiments Bayesian network causal reasoning relations mastery Bayes Theorem, learning Bayesian network causal reasoning

自组织系统Kohonen网络模型。对于Kohonen神经网络，竞争是这样进行的：对于“赢”的那个神经元c，在其周围Nc的区域内神经元在不同程度上得到兴奋，而在Nc以外的神经元都被抑制。网络的学习过程就是网络的连接权根据训练样本进行自适应、自组织的过程，经过一定次数的训练以后，网络能够把拓扑意义下相似的输入样本映射到相近的输出节点上。网络能够实现从输入到输出的非线性降维映射结构：它是受视网膜皮层的生物功能的启发而提出的。~..~-Kohonen network model. For Kohonen neural network, competition is this : For the "winner" of neurons c, in its switching around the region neurons in varying degrees, to be excited, and the switching outside the neurons were inhibited. Network learning is a process in the network connecting the right under the training samples for adaptive, self-organizing process, after a certain number of training, network topology can sense similar to the mapping of the input samples similar to the output nodes. Network can be achieved from input to output of nonlinear reduced-dimensional mapping structure : it is subject to retinal cortex of the biological function inspired by. ~ ~ ..

内附两个源代码，其中一个源代码是识别程序的，另一个是一个矩阵类库的，所以一并附上。 ~..~因为已经保存了训练好的网络权值，所以第一次使用识别程序时，可以直接运行Release目录下的可执行文件，然后对图片目录中的测试图片进行读入、识别。 当然读者也可以自行用训练样本训练网络，不过要特别注意训练样本的选择，否则可能造成识别率很低。训练样本选择的原则是，尽可能的有代表性，在训练时间不至于太长的情况下，训练样本数目尽量多。 -containing 2 source code, which is a source identification procedures, and the other is a matrix class library. So accompanied. ~ ~ .. Has been preserved because of the trained network weights, the first use of the identification procedures, can be run directly Release directory executable file, Then the catalog of tests conducted read into the picture, recognition. Certainly readers can use their own training samples training network, but special attention should be paid to the choice of training samples, Otherwise, it could cause a very low recognition rate. Training sample selection principle is that as far as possible be representative, in the training of the time not too long under the circumstances, the number of training samples as much as possible.