|
From: | Carlo Rossi |
Subject: | kclassify problem |
Date: | Fri, 22 May 2009 18:05:27 +0000 (GMT) |
Hello, I post it here even it's for matlab but maybe it could be interesting as well. I have to use kclassufy to classify a test set of 10 digits (0..9); then I have 10 differents classes. And I have a very large training and test set. I tried to do it with that seems classify in the right way but it takes more than 10 minutes. Is there any mistake or it's simply wrong this way? %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %I fill TRAIN and TEST with only '1' to fill they quickly but it's not real.... TRAIN = [ repmat(1, 50000, 500) ]; target= [repmat(0, size_te/10, 1); repmat(1, size_te/10, 1) ; repmat Kclassification = knnclassify(TEST, TRAIN, group, 1, 'euclidean',
Unfortunately I have a larger test set.......
|
[Prev in Thread] | Current Thread | [Next in Thread] |