摘要
作为模式识别的感知器神经网络,构建网络模型需要一个很重要的权值参数,可以通过手工计算,也可以通过MATLAB神经网络工具箱进行训练仿真获得,使用两种权值建立分类器网络模型,对比测试模式识别结果,不同权值模型对于测试样本识别率是有差异的,为了提高模型的泛化,构建网络模型时需要测试选用最优权值。
As the perceptron neural network pattern recognition,constructing the network model takes a very important parameters,which can be used by manual calculation,and can also carry out the training simulation obtained by MATLAB neural network toolbox. The paper uses two weights for building the classifier network model,and compares test results of the pattern recognition model. It is known that for the recognition rate of the samples tested,the weights are different,and in order to improve the generalization of the model,the construction of the network model need to test and select the optimal weights.
出处
《智能计算机与应用》
2015年第3期93-95,共3页
Intelligent Computer and Applications
关键词
模式识别
神经网络
感知器
权值
Pattern Recognition
Neural Network
Perceptron
Weights
