|
View:
New views
2 Messages
—
Rating Filter:
Alert me
|
|
|
Weka Online - welcome!Hello all.
We have developped a free web-based interface to Weka, the machine-learning suite of Waikato University. Weka Online is available at http://www.ceodelegates.com. Site links on the left tell you more about its features, but ultimately, with a single run of Weka Online, you get all (or most) Experimenter, Explorer and commandline Weka interfaces get. We would appreciate if you keep your datasets small for a while. (There are sample small ARFFs provided on the submission page now.) We are presently migrating to a supercluster of thousands of nodes and a very large Xmx so this request is not permanent and we're prepared to take on as many users as there are (in fact will make it so). We are also working on sending an email notification when your project is ready. Till then, wait a fair time after hitting Send till you go to the Results page. Using commandline Weka, it is only faster than Weka GUI but with current -Xmx1600M it will be relatively slow (till we get migrated). The same site will also be home to CEO ('clever ensembling operator') algorithm. Early versions of it shown to provide +2% to +4% accuracy gain over best single classifier. We are announcing developments on this and probably other mailing lists for a while now, hope no one feels there are too many. We will ultimately have a mailing list that reduces these launch messages. To get in that mailing list, there's two ways: submit a dataset with email address or contact us at developpers@.... We appreciate any comments you have at this early stage. best wishes, ----------------------------------- founders of the Weka Online & CEO, Harri M.T. Saarikoski, designer harri.saarikoski@... Richard J.B. Dobson, programmer richard.dobson@... ----------------------------------- _______________________________________________ Wekalist mailing list Wekalist@... https://list.scms.waikato.ac.nz/mailman/listinfo/wekalist |
|
|
|
| Free Forum Powered by Nabble | Forum Help |