MATLAB Answers

Back Propagation Accuracy improvement

3 views (last 30 days)
Vasumathi Ganesh
Vasumathi Ganesh on 26 Jun 2015
Answered: hatem on 17 Jan 2017
I have used BPN to classify my SAR images.I have done this work and finally performed accuracy also.i got 90% of accuracy. To perform this accuracy i used true positive,true negative,false positive and false negative concept. I have given my code which i used for BPN classification.
A =[5.339 6.1692 5.9526 10.1191 2.3506 0.8338]*100;( this is my image pixel value i performed pixel based classification) x =[0 1 2 3 4 5]; net1 = newff(minmax(A),[60,6,1],{'tansig','tansig','purelin'},'trainrp'); net1.trainParam.show = 1000; net1.trainParam.lr = 0.04; net1.trainParam.epochs = 700; net1.trainParam.goal = 1e-5; [netan] = train(net1,A,x); save abc y = round(abs((sim(netan,A))))
can anyone Please tell me how to improve the accuracy of neural network.
Thank you

  0 Comments

Sign in to comment.

Accepted Answer

Greg Heath
Greg Heath on 27 Jun 2015
Why aren't you using the classification nets patternnet(special form of feedforwardnet) or newpr (special form of newff) with one hidden layer?
Typically, the only thing you have to specify is the number of hidden nodes.
Search the NEWSGROUP and ANSWERS for my classification/pattern-recognition posts.
Hope this helps.
Thank you for formerly accepting my answer
Greg

  1 Comment

Vasumathi Ganesh
Vasumathi Ganesh on 29 Jun 2015
Thank you so much Greg. I will work on this.

Sign in to comment.

More Answers (1)

hatem
hatem on 17 Jan 2017
Vasumathi Ganesh, can ypu post your code ?
Regards

  0 Comments

Sign in to comment.

Sign in to answer this question.