Readme

This package is used to perform feature selection based on phylogenetic beta diversity of microbial community. It takes the advantages of tree structure and branch length to regulate OTUs and higher taxa level. 
The algorithm is designed to solve 5 limitations of routine feature selection problem mentioned in Knights' review paper in 2010 and 2013. The algorithm rely on sparse linear programming and non-linear programming with linear constrain. This matlab version uses software tomlab to speed up calculation (Original matlab optimization tool works a little bit slower). The python SGD implementation for online learning is not included here. 
In order to evaluate and compare this algorithm with others, the evaluation codes rely on 3rd part open source packages and contributions like: libsvm, importCSV, random forest, glmnet... They are not included in main algorithms, but please contact the authors for liscence.