Linear Classification with Online Adaptation of Coordinate Frequencies

Solving the linear SVM problem with coordinate descent is very efficient and is implemented in one of the most often used packages, 'LIBLINEAR' (available at http://www.csie.ntu.edu.tw/~cjlin/liblinear). It has been shown that the uniform selection of coordinates can be accelerated by using an online adaptation of coordinate frequencies (ACF). This package implements ACF and is based on 'LIBLINEAR' as well as the 'LiblineaR' package (< https://cran.r-project.org/package=LiblineaR>). It currently supports L2-regularized L1-loss as well as L2-loss linear SVM. Similar to 'LIBLINEAR' multi-class classification (one-vs-the rest, and Crammer & Singer method) and cross validation for model selection is supported. The training of the models based on ACF is much faster than standard 'LIBLINEAR' on many problems.


News

Reference manual

It appears you don't have a PDF plugin for this browser. You can click here to download the reference manual.

install.packages("LiblineaR.ACF")

1.94-2 by Aydin Demircioglu, 4 years ago


http://github.com/aydindemircioglu/liblineaR.ACF/


Browse source code at https://github.com/cran/LiblineaR.ACF


Authors: Aydin Demircioglu <[email protected]>; Tobias Glasmachers <[email protected]>; Urun Dogan <[email protected]>


Documentation:   PDF Manual  


GPL-2 license


Suggests SparseM, testthat


See at CRAN