Data scaling in neural network
2 views (last 30 days)
Show older comments
mustafa alnasser
on 27 Sep 2015
Answered: Madhav Rajan
on 1 Oct 2015
Dear All;
I would like to know where MATLAB doing the input scaling using MAPMINMAX function , to change the scaling. Suppose that i have the standard code for Classification problem below , does this code do scaling between [-1,1]?:
clc; clear; close all; [x1,TXT,RAW]=xlsread('finalall.xlsx','Ln(3P)'); [t1,TXT2,RAW2]=xlsread('finalall.xlsx','OUT2'); x=x1'; t=t1'; size(x) size(t)
net= patternnet ([40]); % net.divideParam.trainRatio = 70/100; % net.divideParam.valRatio = 15/100; % net.divideParam.testRatio = 15/100; % view(net) net=init(net); [net,tr] = train(net,x,t); nntraintool
plotperform(tr)
1 Comment
Greg Heath
on 28 Sep 2015
Edited: Greg Heath
on 28 Sep 2015
PLEASE LEARN HOW TO FORMAT YOUR POST!
40 hidden nodes is probably much higher than necessary. Did you try using the default (10) first, and, if insufficient, search for the smallest value that works?
The ratio assignments are default values. Therefore they can be omitted.
Accepted Answer
Madhav Rajan
on 1 Oct 2015
I understand that you want to know where the mapminmax scaling is applied when training the network. The 'patternnet' function calls the 'feedforwardnet' function which calls the 'mapminmax' function. Those three functions are MATLAB scripts which can be opened using the 'edit' command. Opening these files and placing breakpoints helped me isolate which function called the 'mapminmax' function.
Hope this helps.
0 Comments
More Answers (0)
See Also
Products
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!