神经网络方法用于分辨3种化学物质

神经网络方法用于分辨３种化学物质冯伟，胡上序（浙江大学化工系，杭州，３１００２７）关键词模式分类，神经网络，模拟退火，遗传算法，传感器阵列传感器阵列技术是利用传感器阵列所提供的交叉敏和模式识别及微机处理技术，来提高传感器的选择性和传感器的测量精度［１...

A Neural Network Algorithm for Pattern Classification of Three Chemical Substances

A hybridized algorithm of artificial neural network(ANN)was used to carry out the task of pattern classification on data from Taguchi Tin-oxided sensors for the three substances acetone,ethanol and methanol.It made use of the Gradient Descent(GD)method with Simulated Annealing(SA)and Genetic algorithm(GA)to speed up the calculation and improve the classification accuracy.It performed good to elude local minima,and consequently improved the accuracy of classfication. Fig.1 shows the ANN learning procedure.The calculation results between two comparison methods are shown in Fig.2 and in Fig.3 for continuing training. By combining the advantages of both SA,GA and GD,extensive tests show that this is a better approach as a pattern recognition method to construct Artificial Nose.