Eigenface和Fisherface表现在姿势

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.

提供了遗传算法的一些代码库文件，国外下的很有用-provide the genetic algorithm code of the document, the foreign useful

文件中包含某老师关于人工神经网络控制系统的PPT，较详细的介绍了人工神经网络及应用于控制的相关内容。-File contains a teacher on the artificial neural network control system of PPT, a more detailed introduction to the artificial neural network and applied to the control of relevant content.

自适应变步长定向变异遗传算法解函数优化问题 需要调用GAOT5.-Adaptive Variable step directional variation of genetic algorithm optimization solutions function call GAOT5 needs.

Mining sequential patterns with constraint (SPADE with constraint)

专家系统课程设计,用计算机解决SP先生谜题,有详细代码-curriculum design expert system, computer SP Mr. Mystery solved a detailed code

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.

采用“提前停止”方法提高 BP 网络的推广能力。对于和例 2相同的问题，在本例中我们将采用训练函数 traingdx -The use of

遗传算法，包含选择，交叉，变异等操作，可求出Y=sin(x)在0-2π的最大值和最小值-Genetic algorithm, including selection, crossover and mutation operation, etc., can be obtained Y = sin (x) at the 0-2π Maximum and minimum

遗传算法中的新的多点交叉算法用来优化选择最优值-Genetic algorithm in the new multi-point crossover algorithm