First test of SVM

[GoogleCodeExporter]

Hi,

I know you must think that i work 24/7 on encog but in fact right now i do.

I have tested your implementation of SVM (i have already tried to integrate 
LibSVM in our datamining tool but it was not finished).

I have found a bug in SVMTrain for the classification problem.

In the constructor you have a loop. In this loop you recreate the svm_problem 
array each time so you end up with an array with null value except for the last 
one.

Easy correction just put the array creation before the loop.

I will keep you informed (probably on the forum) if i have any comments about 
the result for SVM.

Regards.

Julien Blaize

Original issue reported on code.google.com by julien.b...@gmail.com on 13 Oct 2010 at 2:56

[GoogleCodeExporter]

Original comment by heatonre...@gmail.com on 15 Oct 2010 at 1:50

• Changed state: Verified

[GoogleCodeExporter]

Edited the wrong issue.  reopening

Original comment by heatonre...@gmail.com on 15 Oct 2010 at 1:51

• Changed state: Accepted

[GoogleCodeExporter]

Thanks for all of the reports.  I fixed this one.  Also, hope to eventually 
have a more integrated way than copying Encog data to LIBSVM data.

Original comment by heatonre...@gmail.com on 15 Oct 2010 at 2:06

• Changed state: Verified