Wolfram Library Archive

Courseware Demos MathSource Technical Notes
All Collections Articles Books Conference Proceedings
Title Downloads

machine learning framework 1.3 for Mathematica

Thomas Natschlager
Mario Drobics
URL: http://www.unisoftwareplus.com/products/mlf/
Felix Kossak

2005 Wolfram Technology Conference
Conference location

Champaign IL

The machine learning framework for Mathematica is a collection of powerful machine learning algorithms integrated into a framework for the purpose of data analysis. Fuzzy logic is one of its key techniques. The framework allows for combining different machine learning algorithms to solve one single problem. The algorithms are highly parameterizeable. Given this parameterizeability combined with the efficient core engine—realized in C++—of the machine learning framework for Mathematica, the user is able to look at the data with changed parameters in real time. The machine learning framework for Mathematica combines a large variety of distinct algorithms in an optimized computational kernel and the manipulation, descriptive programming, and graphical capabilities of Mathematica give the user unforeseen insights into its data.

Driven by the needs of our industrial partners the machine learning framework is continuously developed to be able to create models from data that generate new knowledge. The upcoming release of the machine learning framework (Version 1.3) introduces:

New algorithms for numerical approximation problems
  • Fuzzy regression trees (FS-LIRT)
  • Additive regression (stochastic gradient boosting)
  • Ridge regression (automatic regularization)
  • Optimized fuzzy rule bases (LAPOC)
New visualization tools
  • Prediction matrix
  • Mutual information matrix
Enhanced model handling
  • Unified training, visualization, testing, storage, and loading of models
Weka interface
  • A J/Link-based interface to Weka, “the Waikato Environment for Knowledge Analysis”

*Wolfram Technology > Application Packages > Applications from Independent Developers > machine learning framework

NatschlaegerDrobicsKossak.zip (918.3 KB) - ZIP archive