Description
The package SAM targets at high dimensional classification
problem for complex data analysis. SAM is short for sparse
additive machine, which is a explicit combination of high
dimensional sparse additive modeling and support vector
machine. Different from existing non-linear classification
methods, which usually use, SAM adopts the computationally
efficient basis spline technique. The optimization is solved by
the Linearized Alternative Direction Method of Multipliers
(L-ADMM). The computation is further accelerated by warm-start
and active-set tricks. For users who are interested in
large-scale problems, we also provide an implementation of L1
norm SVM for computational convenience.