The Altair Community is migrating to a new platform to provide a better experience for you. In preparation for the migration, the Altair Community is on read-only mode from October 28 - November 6, 2024. Technical support via cases will continue to work as is. For any urgent requests from Students/Faculty members, please submit the form linked here
SVM serious performance issues
Learner III
Hi everyone,
i am having serious performance issues with SVM(LibSVM) for one-class learning problems. I was blaming my data so i've decided to try it on some easy stuff, i simplifed the experiment and used self-genereted data, just to see how that goes, same crappy results. I am using RM 5.2 and here's my simplifed experiment:
My "data" consist of 1 integer attribute and 1 nominal label. Label holds an information to which class does this example belong to. My training data consist of data of ONLY 1 CLASS. I have 5 examples in my training data with integer att form 4950 to 5500(pretty tight cluster i think), label set to class1. My testing data have 3 examples: att:4850 no label set, att: 5200 no label set, and 11000 no label set. Of course, desired output is: 4850 belongs to class 1, 5200 belongs in class 1, while 11000 belongs to class 2. Output i am getting is CLASS 2 for all for rbf kernel, and for linear and the rest 5200-1, 4850-2, 1100-1. All parameters left to default(I've tried some modification, no use) I feel that this is a very simple problem to solve, and i cant figure out why it isn't working.
If you guys have some advice i will really appriciate it
i am having serious performance issues with SVM(LibSVM) for one-class learning problems. I was blaming my data so i've decided to try it on some easy stuff, i simplifed the experiment and used self-genereted data, just to see how that goes, same crappy results. I am using RM 5.2 and here's my simplifed experiment:
My "data" consist of 1 integer attribute and 1 nominal label. Label holds an information to which class does this example belong to. My training data consist of data of ONLY 1 CLASS. I have 5 examples in my training data with integer att form 4950 to 5500(pretty tight cluster i think), label set to class1. My testing data have 3 examples: att:4850 no label set, att: 5200 no label set, and 11000 no label set. Of course, desired output is: 4850 belongs to class 1, 5200 belongs in class 1, while 11000 belongs to class 2. Output i am getting is CLASS 2 for all for rbf kernel, and for linear and the rest 5200-1, 4850-2, 1100-1. All parameters left to default(I've tried some modification, no use) I feel that this is a very simple problem to solve, and i cant figure out why it isn't working.
If you guys have some advice i will really appriciate it
Tagged:
0
