tag:blogger.com,1999:blog-30819994.post2994649278494027586..comments2023-05-02T04:56:11.969-04:00Comments on Niketan Pansare's blog: Cross-validation on SVM-LightNiketanhttp://www.blogger.com/profile/04265098294100409965noreply@blogger.comBlogger7125tag:blogger.com,1999:blog-30819994.post-47426319400303616452014-01-08T15:40:16.407-05:002014-01-08T15:40:16.407-05:00Just to confirm, in your MyResults.txt file, entri...Just to confirm, in your MyResults.txt file, entries after "Polynomial SVM with default parameters" are looking good. Also, I suspect you might not be getting entries after the line "RBF SVM with default parameters" except "gamma: 0.00001", "gamma: 0.0001" and so on. Am I correct ? If yes, your shell is probably having problem interpreting the variable "g" in line 46. Try putting a floating point operation (like addition or something) after line 48 and see if we are heading in right direction.Niketanhttps://www.blogger.com/profile/04265098294100409965noreply@blogger.comtag:blogger.com,1999:blog-30819994.post-24327259313824580492014-01-08T13:54:27.334-05:002014-01-08T13:54:27.334-05:00SVM itself is good. I could use it perform default...SVM itself is good. I could use it perform default training and testing.<br />"./svm_light_cross_validation ../feature_set 200 36 10 MyResults.txt" is how I run it, in SVM folder.Anonymoushttps://www.blogger.com/profile/07578070262122469709noreply@blogger.comtag:blogger.com,1999:blog-30819994.post-70930591519167345792014-01-08T13:47:34.450-05:002014-01-08T13:47:34.450-05:00Have you downloaded svmlight and placed svm_learn ...Have you downloaded svmlight and placed svm_learn and svm_classify binaries in the same folder. Also to double check please run "./svm_learn" to see if the binaries are working on your OS.Niketanhttps://www.blogger.com/profile/04265098294100409965noreply@blogger.comtag:blogger.com,1999:blog-30819994.post-87084971592249537932014-01-08T13:28:12.558-05:002014-01-08T13:28:12.558-05:00I might not make me well understood. I have only o...I might not make me well understood. I have only one file to do the experiment. Also, I placed your script under the sum_light folder which is at the same level as my file, errors like "./svm_light_cross_validation: line 46: 3465 Floating point exception: 8 ./svm_learn -t 2 -g $g $TRAIN_FILE $MODEL_FILE > $TEMP_FILE" repeated. can you shed me some light to make it work?Anonymoushttps://www.blogger.com/profile/07578070262122469709noreply@blogger.comtag:blogger.com,1999:blog-30819994.post-56914232864734741762014-01-08T12:23:23.982-05:002014-01-08T12:23:23.982-05:00i used this script for testing it on handful of fi...i used this script for testing it on handful of files. If you have tons of data files to test, I would either change the variables starting from line 7 (eg: MODEL_FILE="./someDir/model."$RANDOM".txt" and so on) or simply write a wrapper shell script over it :)Niketanhttps://www.blogger.com/profile/04265098294100409965noreply@blogger.comtag:blogger.com,1999:blog-30819994.post-47786700299250405152014-01-08T11:00:52.135-05:002014-01-08T11:00:52.135-05:00Thanks for your instructions. But in which folder,...Thanks for your instructions. But in which folder, should I run the svm_light_cross_validation script?Anonymoushttps://www.blogger.com/profile/07578070262122469709noreply@blogger.comtag:blogger.com,1999:blog-30819994.post-88150799297695837622012-05-07T10:49:37.801-04:002012-05-07T10:49:37.801-04:00This is a great contribution. My honest thanks.This is a great contribution. My honest thanks.Freddy Lopezhttps://www.blogger.com/profile/04560516889321267412noreply@blogger.com