Azzera filtri
Azzera filtri

problems with training function LVQ (example from help does not work)

2 visualizzazioni (ultimi 30 giorni)
Hi all, please I need your help !
example (LVQ) from help does not work
P = [-3 -2 -2 0 0 0 0 2 2 3; 0 1 -1 2 1 -1 -2 1 -1 0];
Tc = [1 1 1 2 2 2 2 1 1 1];
T = ind2vec(Tc);
targets = full(T);
net = newlvq(P,4,[.6 .4]);
Y = sim(net,P);
Yc = vec2ind(Y);
net.trainParam.epochs = 150;
net = train(net,P,T);
In training process after 2 interactions training was stop. (My system 2011a,a check on 2009b - the same thing).
Maybe I'm doing something wrong?

Risposta accettata

Greg Heath
Greg Heath il 2 Ott 2011
Nothing is wrong...it is a simple problem that only requires 2 or 3 epochs to obtain 0 error. You just didn't follow through.
close all, clear all, clc
% NOTE: Remove end semicolons to monitor outputs
P = [-3 -2 -2 0 0 0 0 2 2 3; 0 1 -1 2 1 -1 -2 1 -1 0]
Tc = [1 1 1 2 2 2 2 1 1 1]
T = ind2vec(Tc)
targets = full(T)
net = newlvq(minmax(P),4,[.6 .4]); % Old Version
% With no training
IW = net.IW{1,1}
LW = net.LW{2,1}
b = net.b(1,:)
Y = sim(net,P)
Yc = vec2ind(Y)
% ... Yc = [ 1 1 1 1 1 1 1 1 1 1 ]
% resulting in a classification error of 40%
%Training with defaults
net = train(net,P,T);
% Converges in 3 epochs
IW = net.IW{1,1}
LW = net.LW{2,1}
b = net.b(1,:)
Y = sim(net,P)
Yc = vec2ind(Y)
% ... Yc = [ 1 1 1 2 2 2 2 1 1 1 ]
% resulting in a classification error of 0% !!!
Hope this helps.
Greg

Più risposte (1)

Greg Heath
Greg Heath il 2 Ott 2011
Sorry for the formatting. This text editor sucks.
Greg
  1 Commento
Walter Roberson
Walter Roberson il 2 Ott 2011
Put a newline before the code to be formatted. Highlight the code to be formatted. Click on the "{} Code" button. Code will now be formatted in the preview.

Accedi per commentare.

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by