A question about training images using ANN

1 次查看（过去 30 天）

Ampi 2012-11-24

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/54685-a-question-about-training-images-using-ann

Hello,

I am facing a problem with training of images. I have a 2D matrix where each row represents a class and in this manner I have 20 classes and each class contains only 1 image of a person. Now I have the final training matrix consisting of 20 rows and 20 columns out of which I have chosen 10 most distinguished feature vectors.So, both my training-matrix consists of 20X20 dimensions and my target matrix contains 20X20 dimensions. Now my problem is, while training using newff function, I have set the following parameters:-

 net.trainParam.epochs=200; 
 net.trainParam.goal=3e-5; 
 net.trainParam.show=50 ;  
 net.trainParam.lr=0.06;  
 net.trainParam.passes=99;

The problem is if I increase the no of distinguished vectors by 30/40, the training is stopping at 13 iterations. But, how can I train my samples for say(200 iterations)? And is the parameters set by me above all right for 20X20 vectors because I am not getting a very good recognition rate? Kindly help.