Genetic algorithm to initialize the network weights and threshold parameters imp...

数据挖掘中很不错的源码,一个share算法的java源程序,有极好的可读性-data mining is a good source, a share algorithm java source files, has an excellent readability

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

人工智能AI经典入门书，很难找的噢，给喜欢的朋友-classical AI Artificial Intelligence portal, it is very difficult to find the, uh, likes to friends

bp神经网络 c语言程序-bp neural network c Language Program

用贪心法解决TSP问题.输入数据可求得所要的最值的结果.-greedy algorithm used to solve TSP. Input data can be obtained by the most value to the results.

详细介绍了SVM 和RVM的区别及各自的优点，说明RVM在很多方面优于SVM-Described in detail the difference between SVM and the RVM and their respective advantages, indicating RVM in many ways superior to SVM

EBP反向传播算法对应的源程序，三层网络，分隐层神经元，输出层神经元，输入层神经元的一个基本的程序实现-EBP back-propagation algorithm the corresponding source code, the three-tier networks, sub-hidden layer neurons, output layer neurons, input layer neurons realize a basic procedure

神经网络讲义课件：PPT格式- Nerve network printed lecture class: PPT form

Hopfield 网――擅长于联想记忆与解迷路 实现H网联想记忆的关键，是使被记忆的模式样本对应网络能量函数的极小值。 设有M个N维记忆模式，通过对网络N个神经元之间连接权 wij 和N个输出阈值θj的设计，使得: 这M个记忆模式所对应的网络状态正好是网络能量函数的M个极小值。 比较困难，目前还没有一个适应任意形式的记忆模式的有效、通用的设计方法。 H网的算法 1）学习模式――决定权重 想要记忆的模式，用-1和1的2值表示 模式：-1，-1，1，-1，1，1，... 一般表示： 则任意两个神经元j、i间的权重： wij=∑ap(i)ap(j)，p=1…p； P：模式的总数 ap(s)：第p个模式的第s个要素（-1或1） wij：第j个神经元与第i个神经元间的权重 i = j时，wij=0，即各神经元的输出不直接返回自身。 2)想起模式： 神经元输出值的初始化 想起时，一般是未知的输入。设xi(0)为未知模式的第i个要素（-1或1） 将xi(0)作为相对应的神经元的初始值，其中，0意味t=0。 反复部分：对各神经元，计算： xi (t+1) = f (∑wijxj（t）-θi), j=1…n, j≠i n―神经元总数 f（）--Sgn（） θi―神经元i发火阈值 反复进行，直到各个神经元的输出不再变化。-Hopfield network-- good at associative memory solution with the realization of lost H associative memory networks, are key to bringing the memory model samples corresponding network energy function of the minimum. With M-N-dimensional memory model, the network N neurons connect between right wij and N output threshold j design makes : M-mode memory corresponding to the network is a state network energy function is the M-000 minimum. Mor

The purpose of this computer program is to allow the user to construct, train and test differenttypes of artificial neural networks. By implementing the concepts of templates, inheritance andderived classes from C++ object oriented programming, the necessity for declaring multiple largestructures and duplicate attributes is reduced. Utilizing dynamic binding and memory allocationafforded by C++, the user can choose to develop four separate types of neural networks: