SPARE (Something for PAttern REcognition)

The SPARE library (Something for PAttern REcognition) is a set of C++ (mainly template) classes offering some building blocks to create software modules for Soft Computing and Pattern Recognition tasks, as: Classification, Clustering, Prediction, Function Approximation, building of Fuzzy models. SPARE is an header-only library, without any need to link (almost) anything, everything is done at compile time. The philosophy behind the SPARE design is to provide simple meta-algorithms. That is, some classical machine learning routines like clustering and genetic algorithms are implemented in a highly flexible and generic way, with the aim of allowing a broad variety of operating scenarios. The SPARE classes do simple things in an effective way, they try to capture the essence of the algorithms. The various template classes can be specialized and combined to build complete algorithms. In order to guarantee the interoperability, the class interface standardization method known as concept based metaprogramming is adopted. While making things less obvious and guided with respect to approaches based on inheritance, this static polymorphism paradigm allows to avoid late binding related performance losses. As concerns the programming style, SPARE tries to be more compliant as possible to the C++ standard library. Data passing interfaces based on STL-like iterators and containers are adopted. The SPARE library makes use also of the very useful boost libraries. The SPARE library is permanently under development, according to our research agenda. The SPARE library is published as an Open Source Project in SourceForge (http://sourceforge.net/p/libspare/home/Spare/), together with a complete documentation and many tutorial application. The SPARE library has been conceived in order to act as an effective repository for our research team an to increase the international visibility of our activities.