Description: 基于C开发的三个隐层神经网络，输出权值、阈值文件，训练样本文件，提供如下函数：1）初始化权、阈值子程序；2）第m个学习样本输入子程序；3）第m个样本教师信号子程序；4）隐层各单元输入、输出值子程序；5）输出层各单元输入、输出值子程序；6）输出层至隐层的一般化误差子程序；7）隐层至输入层的一般化误差子程序；8）输出层至第三隐层的权值调整、输出层阈值调整计算子程序；9）第三隐层至第二隐层的权值调整、第三隐层阈值调整计算子程序；10）第二隐层至第一隐层的权值调整、第二隐层阈值调整计算子程序；11）第一隐层至输入层的权值调整、第一隐层阈值调整计算子程序；12）N个样本的全局误差计算子程序。-C development based on the three hidden layer neural network, the output weights, threshold documents, training sample documents, for the following functions : a) initialization, the threshold subroutine; 2) m learning samples imported subroutine; 3) m samples teachers signal Subroutine ; 4) hidden layer of the module input and output value subroutine; 5) the output layer of the module input and output value subroutine; 6) the output layer to the hidden layer subroutine error of generalization; 7) hidden layer to the input layer subroutine error of generalization; 8) the output layer to the third hidden layer Weight adjustment, the output layer threshold adjustment routines; 9) 3rd hidden layer to the second hidden layer weights adjustment, the third hidden layer threshold adjustment routi Platform: |
Size: 11264 |
Author:李洋 |
Hits:
Description: 简单的神经网络训练程序，可以通过增加训练样本达到更好的权值-Simple neural network training procedures, increasing the number of training samples can achieve better weight Platform: |
Size: 1024 |
Author:顾添锦 |
Hits:
Description: BP神经网络，是利用人工智能的方法，训练权值，建立模式识别的模型。本程序，采用三层神经网络，对隐层的节点个数使用经验公式，对其他的参数设置也进行了优化，实现了手写数字的识别，-BP neural network is the use of artificial intelligence methods, weight training, the establishment of pattern recognition models. This procedure, using three neural networks, on the number of hidden layer nodes using the empirical formula, on the other parameters are also optimized to achieve a hand-written digit recognition, Platform: |
Size: 263168 |
Author:胡存英 |
Hits:
Description:
％ Train a two layer neural network with the Levenberg-Marquardt
％ method.
％
％ If desired, it is possible to use regularization by
％ weight decay. Also pruned (ie. not fully connected) networks can
％ be trained.
％
％ Given a set of corresponding input-output pairs and an initial
％ network,
％ [W1,W2,critvec,iteration,lambda]=marq(NetDef,W1,W2,PHI,Y,trparms)
％ trains the network with the Levenberg-Marquardt method.
％
％ The activation functions can be either linear or tanh. The
％ network architecture is defined by the matrix NetDef which
％ has two rows. The first row specifies the hidden layer and the
％ second row specifies the output layer.- Train a two layer neural network with the Levenberg-Marquardt method. If desired, it is possible to use regularization by weight decay. Also pruned (ie. not fully connected) networks can be trained. Given a set of corresponding input-output pairs and an initial network, [W1, W2, critvec, iteration, lambda] = marq (NetDef, W1, W2, PHI, Y, trparms) trains the network with the Levenberg-Marquardt method . The activation functions can be either linear or tanh. The network architecture is defined by the matrix NetDef which has two rows. The first row specifies the hidden layer and the second row specifies the output layer. Platform: |
Size: 3072 |
Author:张镇 |
Hits:
Description: Image thresholding has played an important role in image segmentation. In this paper, we present a novel spatially weighted fuzzy c-means (SWFCM) clustering algorithm for image thresholding. The algorithm is formulated by incorporating the spatial neighborhood information into the standard FCM clustering algorithm. Two improved implementations of the k-nearest neighbor (k-NN) algorithm are introduced for calculating the weight in the SWFCM algorithm so as to improve the performance of image thresholding. Platform: |
Size: 293888 |
Author:silviudog |
Hits:
Description: 神经网络是单个并行处理元素的集合，我们从生物学神经系统得到启发。在自然界，网
络功能主要由神经节决定，我们可以通过改变连接点的权重来训练神经网络完成特定的功能-A single neural network is a collection of parallel processing elements, we have been inspired by biological nervous system. In nature, the network function of a decision by the ganglion, we can by changing the weight of the connection point to train neural network to complete the function of specific Platform: |
Size: 244736 |
Author:zgp |
Hits:
Description: 用MATLAB实现二层bp神经网络的计算。可以改变阈值和权值以改进算法，并可以将该方法推广到多层网络。-Using MATLAB to achieve the second floor bp neural network computing. Can change the threshold value and weight to improve the algorithm and the method can be extended to the multi-layer network. Platform: |
Size: 3072 |
Author:lyr |
Hits:
Description: 用遗传算法优化BP神经网络的初始权值与阈值.使用gaot与nntools-Use ga optimative the weight and bias of BP neural network. Platform: |
Size: 35840 |
Author:dynamicansys |
Hits:
Description: SOM神经网络（自组织特征映射神经网络）是一种无导师神经网路。网络的拓扑结构是由一个输入层与一个输出层构成。输入层的节点数即为输入样本的维数，其中每一节点代表输入样本中的一个分量。输出层节点排列结构是二维阵列。输入层X中的每个节点均与输出层Y每个神经元节点通过一权值(权矢量为W)相连接，这样每个输出层节点均对应于一个连接权矢量。
自组织特征映射的基本原理是，当某类模式输入时，其输出层某一节点得到最大刺激而获胜，获胜节点周围的一些节点因侧向作用也受到较大刺激。这时网络进行一次学习操作，获胜节点及其周围节点的连接权矢量向输入模式的方向作相应的修正。当输入模式类别发生变化时，二维平面上的获胜节点也从原来节点转移到其它节点。这样，网络通过自组织方式用大量训练样本数据来调整网络的连接权值，最后使得网络输出层特征图能够反映样本数据的分布情况。根据SOM网络的输出状况，不仅能判断输入模式所属的类别，使输出节点代表某类模式，而且能够得到整个数据区域的分布情况，即从样本数据得到所有数据的分布特征。 -SOM neural network (self-organizing feature map neural network) is an unsupervised neural network. Network topology is an input layer and an output layer. Input layer nodes is the input dimension of the sample, each node represents a component input samples. Output layer nodes are arranged in two-dimensional array structure. X in the input layer and output layer each node of each neuron node Y by a weight (the weight vector as W) is connected, so that each output layer corresponds to a connection node of the right vector.
Self-organizing feature maps of the basic principle is, when each category of inputs into the model, its output layer one node get the maximum boost and win, Huoshengjiedian around Yixiejiedian Yin Zuo Yong Ye Shoudaojiaotai lateral stimulation. Then a learning network operation, the winner node and surrounding nodes in the right direction vector to the input mode to make consequential amendments. When the input mode type changes, the two-dimensional plane of the wi Platform: |
Size: 47104 |
Author:leidan |
Hits:
Description: 实验中使用随机反向传播算法对构造的神经网络进行学习，最终得到构造的神经网络的权值矩阵。-Experiment using the random back-propagation algorithm to construct the neural network learning, the final structure of the neural network obtained the weight matrix. Platform: |
Size: 2048 |
Author:王瑶 |
Hits:
Description: An Improved PSO Algorithm to Optimize BP Neural Network
Abstract
This paper presents a new BP neural network
algorithm which is based on an improved particle swarm
optimization (PSO) algorithm. The improved PSO (which
is called IPSO) algorithm adopts adaptive inertia weight
and acceleration coefficients to significantly improve the
performance of the original PSO algorithm in global
search and fine-tuning of the solutions. This study uses the
IPSO algorithm to optimize authority value and threshold
value of BP nerve network and IPSO-BP neural network
algorithm model has been established. The results
demonstrate that this model has significant advantages
inspect of fast convergence speed, good generalization
ability and not easy to yield minimal local results Platform: |
Size: 252928 |
Author:dasu |
Hits:
Description: Kohonen神经网络算法工作机理为：网络学习过程中，当样本输入网络时，竞争层上的神经元计算输入样本与竞争层神经元权值之间的欧几里德距离，距离最小的神经元为获胜神经元。调整获胜神经元和相邻神经元权值，使获得神经元及周边权值靠近该输入样本。通过反复训练，最终各神经元的连接权值具有一定的分布，该分布把数据之间的相似性组织到代表各类的神经元上，使同类神经元具有相近的权系数，不同类的神经元权系数差别明显。需要注意的是，在学习的过程中，权值修改学习速率和神经元领域均在不断较少，从而使同类神经元逐渐集中。-Kohonen neural network algorithm for the working mechanism: the network learning process, when the samples enter the network, the competitive layer of neurons on the calculation of input samples and competitive layer neurons Euclidean distance between the weights from the smallest neurons Winning neuron. Adjust the winning neuron and neighboring neurons weights to gain weight and peripheral neurons close to the input samples. Through repeated training, and ultimately the connection weights of neurons with a certain distribution, the distribution of the similarity between the data to representatives of organizations of various types of neurons, so that similar neurons have similar weights, different types of nerve Element obviously different weights. Note that, in the learning process, the right to modify the value of the field of learning rate and neurons were constantly low, so that the same neurons gradually concentrated. Platform: |
Size: 87040 |
Author:李芳 |
Hits:
Description: 模拟BP神经网络，已知四个输入向量和期待输出，可以通过hebbian和delta学习算法得到权重值，从而构建网络-BP neural network simulation, the four known input vector and look forward to the output, you can get hebbian and delta weight learning algorithm to build the network Platform: |
Size: 1024 |
Author:孙涛 |
Hits:
Description: 基于遗传算法和BP神经网络控制倒立摆的程序，用遗传算法优化神经网络权值阈值以达到更好的控制效果-Based on genetic algorithm and BP neural network control procedures for the inverted pendulum, with a genetic algorithm neural network weight threshold in order to achieve better control performance Platform: |
Size: 37888 |
Author:李增光 |
Hits:
Description: 神经网络进行权重分析，能够帮助我们做前期分析，可以试试。-Neural network weight analysis, the preliminary analysis can help us, you can try. Platform: |
Size: 7168 |
Author:ljb |
Hits:
Description: 城市交通流的运行存在着高度的复杂性、时变性和随机性,实时准确的交通流预测是智能交通系统,特别是先进的交通管理系统与先进的出行者信息系统研究的关键. 基于交通流预测的特点,给出了基于遗传算法的小波神经
网络的交通预测模型GA-WNN ,用具有自然进化规律的遗传算法来对小波神经网络的连接权值和伸缩平移尺度进行前期优化训练,部分代替了小波框架神经网络中按单一梯度方向进行参数优化的梯度下降法,克服了单一梯度下降法易陷入局部极小和引起振荡效应等缺陷. 仿真实验验证了GA-WNN 预测模型对短时交通流的预测的有效性.-For the high complexity ,time-variation and probability of urban traffic flow , its real-time and exact
prediction is critical to the research of intelligent traffic systems , especially for the advanced traffic manage-ment system and advanced traveler information system. Based on the character of the traffic flow prediction , a
GA-WNN model is given based on the wavelet neural network with genetic algorithm. The genetic algorithm of
natural evolving law for the gradient descendent algorithm in Wavelet Neural Network is partly substituted to
pre-optimize the connection weight and the extension scale of the wavelet neural network and later optimize the
parameters along a single gradient vector. This method overcomes some drawbacks when there exists a single
gradient descendent algorithm , such as local minimum and oscillation. A short-time traffic flow prediction sim-
ulation using the GA2WNN prediction model demonstrates the validity of the model . Platform: |
Size: 615424 |
Author:mengfei |
Hits:
Description: 通过遗传算法优化BP神经网络权系数，达到局部的最有、优-By genetic algorithm to optimize BP neural network weight coefficient to reach the most local, excellent Platform: |
Size: 4096 |
Author:王林 |
Hits:
Description: 基于样条权函数神经网络的手写系统实现。验证了样条权函数神经系统在手写数字识别中的应用价值。-Function neural network system to achieve spline right hand. Verify that the spline weight function of the nervous system applications in handwritten digits recognition. Platform: |
Size: 808960 |
Author:zhaohua |
Hits:
Search - neural network weight