Machine Learning Workbench

With a number of available machine learning workbenches, why create another one? "The need for open source software in machine learning" is carefully laid out in the following work [Sonnenburg, 2007]. In short, the Journal of Machine Learning Research at Massachusetts Institute of Technology has started a new publication track for open source machine learning workbenches in order to provide more incentive to develop such open source software. The advantages of open source software for machine learning research includes:

• Reproducibility of results, leading to fair comparisons
• Uncovering problems in current algorithms
• Building on existing resources

malibu: A Machine Learning Workbench

malibu defines a set of core classifiers that form the foundation of the workbench. In order to strengthen or extend the classifiers to new domains, malibu also defines a set of wrappers. Unlike other workbenches, the classifiers and wrappers are selected at compile time giving maximum efficiency during run-time. malibu utilizes C++ meta programming, which has been made popular by projects like Boost and the The Matrix Template Library. The template mechanism permits high performance modular programming [9].