MATLAB Answers

2

how to save and reuse a trained neural network

Asked by Omotayo Asiru on 18 Jan 2016
Latest activity Commented on by Greg Heath
on 18 May 2019 at 4:31
I just trained a neural network and i will like to test it with new data set that were not included in the training so as to check its performance on new data. This is my code; net = patternnet(30); net = train(net,x,t); save (net); y = net(x); perf = perform(net,t,y) classes = vec2ind(y); where x and t are my input and target respectively. I understand that save net; can be used but my questions are as follows ; 1.At what point in my code will i put save net 2.Using save net;, which location on the system is the trained network saved? 3.How can i load the trained network and supply new data that i want to test it with? Please Note: I want to be able to save the trained neural network such that when i run the code over and over again with the training data set,it gives same output. I have discovered that each time i run my code,it gives a different output which i do not want once i have an acceptable result. Please help

  0 Comments

Sign in to comment.

3 Answers

Answer by Greg Heath
on 26 Jan 2016
 Accepted Answer

1.At what point in my code will i put save net
Any time after training it and before deleting it.
However, give it a unique name so that it is not overwritten
or used by mistake.
gregnet1 = net;
save gregnet1
2.Using save net;, which location on the system is the trained network saved?
What ever directory you are in when you save it UNLESS you
specify another directory.
3.How can i load the trained network and supply new data that i want to test it with?
load gregnet1
newoutput = gregnet1(newinput);
Please Note: I want to be able to save the trained neural network such that when i run the code over and over again with the training data set,it gives same output. I have discovered that each time i run my code,it gives a different output which i do not want once i have an acceptable result.
Then initialize the RNG to the same state before training to
obtain reproducibility. See any of my training example posts.
Hope this helps.
Thank you for formally accepting my answer
Greg

  3 Comments

If you are just loading a trained net and using it on the same data, you should get the same answer.
Somewhere along the line either you net or your data is changing.
Can you tell us which it is?
Greg
Thanks for this explanation. I think my problem has been saving and loading the net because when i use save net,i really can not be sure of the network its loading as i have never used a unique name to save the network. I will work with your explanations above and feed you back on the outcome.
Please, how can i initialize RNG to same state to obtain reproductivity?
how do you specify a different directory to save the the network in ?

Sign in to comment.


Answer by Greg Heath
on 22 Jan 2016

save net ...
...
load net
Hope this helps.
Thank you for formally accepting my answer
Greg

  1 Comment

Thanks for your response but this has not answered my question.As i said in my question,i know you save net and load net can be used but my questions are: 1.At what point in my code will i put save net 2.Using save net;, which location on the system is the trained network saved? 3.How can i load the trained network and supply new data that i want to test it with? Please Note: I want to be able to save the trained neural network such that when i run the code over and over again with the training data set,it gives same output. I have discovered that each time i run my code,it gives a different output which i do not want once i have an acceptable result.

Sign in to comment.


Answer by Ayesha Zafar on 17 May 2019 at 16:51

hello!
I am trying to save and load the net but when i test the net keeping train function in comment.It give error that is "Undefined function or variable 'gregnet1'. " Attached screenshot is explaining the problem
close all, clear all, clc, format compact,
img = imread('nn3.bmp');
% imshow(img);
% imgGray = rgb2gray(img);
% imgCrop = imcrop(imgGray);
% imshow(imgCrop);
% imgLGE=imresize(imgCrop,[7,5 ]);
% imshow(imgLGE);
% imgRTE = imrotate(imgLGE, 35);
% imshow(imgRTE);
% imgBW = im2bw(imgLGE, 0.90455);
% imshow(imgBW);
% imwrite(imgBW,'nine3.bmp');
input = imread('nine.bmp');
corresponding target output vector
output=[ -1 -1 -1 -1 -1 -1 -1 -1 -1 1];
net = network( 1,1,0,1,0,1);
net.layers{1}.size = 10;
net.layers{1}.transferFcn = 'tansig';
net = configure(net,input(:),output(:));
% net = init(net);
view(net);
initial_output = net(input(:))
net.trainFcn = 'traingd'; %the term “backpropagation” is sometimes used to refer specifically to the gradient descent algorithm,
[net,tr]= train(net,input(:),output(:)); %training record= tr
final_output = net(input(:))
gregnet1 = net;
save gregnet1
%test
% output=sim(net,input(:))
load gregnet1
newoutput = gregnet1(input(:))

  1 Comment

  1. After you train net, what are the training,validation and test subset error rates ???
  2. tr = tr % = ?
Greg

Sign in to comment.