by Don Park
at 2005/04/01 00:32:31
Is there a decent open source Java Bayesian package that is not GPL or similarly restricted from commercial use? I am aware of only Classifier4J. Preferably, it should be optimized for server applications and high performance.
Comments
No luck Don. I did a similar search 6 months ago and found only weka and xelopes.
This is my feedback after just reading the docs for the 2 libraries. I haven't actually used them.
From a licensing perspective:
* Weka is university work and GPL, so no way to get a commercial license.
* Xelopes is made by a company and has a dual licensing scheme (GPL to play with/proprietary to deploy for profit), so you could buy a license from them.
From a scalability perspective:
Weka is university type code, designed to work on text files as input. Lots of plumbing required to use it in a highly scalable server side environment.
Xelopes is designed for server application, high performance, production type data mining applications, with some connectors already done for you.
As for Classifier4J, Nick's project at that time was very basic, and documentation was one page. Looking at the javadoc implementation it did not seem very advanced. Not sure it would have bought us a lot. But that was 6 months ago: I'm sure Nick enhanced it since then.
Overall what you ask for is Xelopes, but in order to use it in a commercial product, you'll have to buy a license from them.
Sorry pal: there's no free lunch:-)
This is my feedback after just reading the docs for the 2 libraries. I haven't actually used them.
From a licensing perspective:
* Weka is university work and GPL, so no way to get a commercial license.
* Xelopes is made by a company and has a dual licensing scheme (GPL to play with/proprietary to deploy for profit), so you could buy a license from them.
From a scalability perspective:
Weka is university type code, designed to work on text files as input. Lots of plumbing required to use it in a highly scalable server side environment.
Xelopes is designed for server application, high performance, production type data mining applications, with some connectors already done for you.
As for Classifier4J, Nick's project at that time was very basic, and documentation was one page. Looking at the javadoc implementation it did not seem very advanced. Not sure it would have bought us a lot. But that was 6 months ago: I'm sure Nick enhanced it since then.
Overall what you ask for is Xelopes, but in order to use it in a commercial product, you'll have to buy a license from them.
Sorry pal: there's no free lunch:-)
NO FREE LUNCH? WAAAAHHH~
It's odd that most Java Bayesian projects are dominantly GPL-based.
Anyhoo, looks like I am just going to experiment with Classifier4J and then roll my own.
It's odd that most Java Bayesian projects are dominantly GPL-based.
Anyhoo, looks like I am just going to experiment with Classifier4J and then roll my own.
Mail me if you've got questions on Classifier4J (nlothian at apache dot org)
Thanks, Nick, for the offer to help and for writing Classifier4J. For now, I don't have any questions. It looks pretty simple and straight forward to work with. My application will use a relatively small word set (50-200) so I think I'll just use a HashMap for WordSource and write it out in textual form (XML or CSV).
Comment has been disabled for this post.
Copyright © 2006 Don Park
