My aim is to generate a training set. So, I have text file including number of images, I select randomly some images and I would like to save them in an other text file.
Can someone help me to correct my MATLAB source code?
noOfImage=176;
trainingSet1= 10;
P1 = myrandint(trainingSet1,1,[1:noOfImage],'noreplace')
P1 = sort(P1);
inputFileID1 = fopen('malin.txt','rt')
while ~feof( inputFileID1)
tline1=fgetl(inputFileID1)
outputFileID = fopen('train.txt','wt');
count1 = 0;
for i=1:trainingSet1
for j=count1+1:noOfImage
count1 = count1+1;
if count1
Hi Khadidja,
In the following matlab program, I have read a txt file which I have attached herewith. Then the elements in the file have been read and saved into a cell array. Then a random array has been created. You can specify your own random array. But the number of elements in that random number array should be less than the number of elements available in you cell array (matin), in which you have all the objects read from the input txt file.
You can change the following code accordingly to suit your problem.
---------------------------------------------------------------------------------------------------
fid = fopen('E:\input.TXT')%You may change the path accordingly
foutFile=fopen('E:\output.TXT','w')
increment=1;
matin{:,1}=[];%matin is a cell matrix, I have taken it as a cell matrix
%because the size of the each object in the input file may vary
while ~feof(fid)
x = fgetl(fid);
matin{increment,1}={x(1,:)};
increment=increment+1;
end
randnum=randi([1, size(matin,1)]); %Here I select only one random number
%between 1 and the size of matin.
%You can define an array of random numbers, but the size of that random
%array must be less than number of objects available in matin
for i=1:size(randnum,1)%
valforout=matin{randnum(1,:),1};%Randomly selected value to be saved
fprintf(foutFile,char(valforout));%Saving randomly selected values in
%the file
fprintf(foutFile,'\r\n');
end
fclose(fid);
fclose(foutFile);
---------------------------------------------------------------------------------------------------
Regards.
Dear Friend,
Please be carefull if you want to use NN in matlab. It makes bias and weight vecvtor start randomly so the performance you obtain every run will be different. Therefore, you will have different performances for each training and testing procedure. To overcome this, make these vectors start from a value you already define. Okan
Dear Khadidja,
I'm sending fere enclosed a Matlab routine for random subdivision into train and test set through stratified sampling. Hope it helps.
Best,
f.
- Badges
- Science topic