Software
From: William H. Hsu
Date: Thu, 9 May 2002 02:17:56 -0500
Subject: MLJ (Machine Learning in Java) 1.0a, open source suite of Java Tools
I am writing to announce the first release of Machine Learning in
Java (MLJ), an open-source suite of Java tools for research in machine
learning.
MLJ currently includes the following modules:
- Java port of MLC++ base inducer and categorizer classes
- Decision tree inducers: ID3, C4.5
- Simple (Naive) Bayes, discrete - improvements cf.
[Kohavi, Becker, and Sommerfield, 1996]
- Wrappers for variable selection
* Search-based Feature Subset Selection (FSS)
[Kohavi and John, 1997]
* CHC genetic algorithm
[Eshelman, 1990; Guerra-Salcedo and Whitley, 1999]
The alpha release of MLJ 1.0 (08 May 2002 build) shall shortly be
available on Sourceforge and for download with or without Javadoc. MLJ
aims toward:
- Full MLC++ [Kohavi et al, 1997] and WEKA 3
[Witten and Frank, 2002] compatibility
- Command-line and configuration file-based control of
high-level
parameters (for ease of experimentation on high-performance
computing systems), cf. ECJ [Luke, 2001]
A full beta release scheduled for 08 Aug 2002 shall also include
interconversion tools among many data formats.
For more information, please see:
http://www.kddresearch.org/Groups/Machine-Learning/MLJ
MLJ is governed by the GNU Public License.
-Bill
| |
|