A wrapper around the 'LIBLINEAR' C/C++ library for machine learning (available at < http://www.csie.ntu.edu.tw/~cjlin/liblinear>). 'LIBLINEAR' is a simple library for solving large-scale regularized linear classification and regression. It currently supports L2-regularized classification (such as logistic regression, L2-loss linear SVM and L1-loss linear SVM) as well as L1-regularized classification (such as L2-loss linear SVM and logistic regression) and L2-regularized support vector regression (with L1- or L2-loss). The main features of LiblineaR include multi-class classification (one-vs-the rest, and Crammer & Singer method), cross validation for model selection, probability estimates (logistic regression only) or weights for unbalanced data. The estimation of the models is particularly fast as compared to other libraries.
Note: In this file, LiblineaR refers to this R package, while LIBLINEAR refers to the original C/C++ library.
1.94-2: 2015/01/30 1. Cleaned DESCRIPTION
2. library(SparseM) replaced by 'require' or 'requireNamespace' as appropriate.
1.94-1: 2015/01/30 1. Upgraded to LIBLINEAR version 1.94, including support vector regression - argument labels of function LiblineaR was renamed as target. The old naming is still accepted with a warning - argument type of function LiblineaR may take extra values (11,12,13) - extra argument svr_eps of function LiblineaR for the tolerance of regression loss - if class labels are -1 and 1, ensure that positive decision values represent the class 1 - enriched examples, with cases of regression
2. Fix CITATION file format to satisfy CRAN requirements 3. Cleaning and reformatting of man pages 4. Fixing a bug in class labels ordering
1.80-11: 2014/09/18 1. Refactored C code in order to facilitate upgrades of LIBLINEAR.
2. Fix memory leak in predictLinear (the buffer x wasn't freed)
1.80-10: 2014/09/14 1. Shortened examples runtime.
2. Uniformized notations in DESCRIPTION file for LiblineaR (this package) and LIBLINEAR (the C/C++ library wrapped by this package).
1.80-9: 2014/09/12 1. Replaced all expressions:
rand()%(a-b); by GetRNGstate(); (int) (unif_rand()*(a - b))%(a - b); PutRNGstate(); in linear.cpp, where a and b might take different values.
1. Corrected a bug in memory allocation for sparse matrices: src/trainLinear.c These changes are to be credited to Christian Wolf. 2. Added a PACKAGE argument to speed up call to .C(...): R/LiblineaR.R R/predict.R
1. Modification of the following files in order to support sparse matrices: R/LiblineaR.R R/predict.R src/predictLinear.c src/trainLinear.c These changes are mainly to be credited to Kai-Hsiang Hsu, from the Department of Computer Science of the National Taiwan University. 2. Addition of examples to reflect the use of sparse matrices in the following file: man/LiblineaR.Rd
1. Corrected a memory mapping bug in predictLinear.c
1. Suppress printing to stdout in linear.cpp, tron.cpp 2. Suppress the use of exit(1); in predictLinear.c 3. Correct a bug when retrieving weights from C to R for multi-class models
1. Correct bugs in linear.cpp (update of solve_l1r_l2_svc and solve_l1r_lr)
1. Add: extern "C" before each function below : // // Interface function // in linear.cpp 2. Suppress useless functions: - save_model - load_model in: linear.cpp linear.h 3. change all "fprintf" into Rprintf in : linear.cpp trainLinear.c predictLinear.c 4. change: #define Malloc(type,n) (type *)malloc((n)*sizeof(type)) into #define Malloc(type,n) (type *)Calloc(n,type) in: trainLinear.c linear.cpp 5. replace all the malloc((n)*sizeof(type)) by Calloc(n,type) in predictLinear.c 5. replace all the realloc(*p,size) by Realloc(*p,n,type) in linear.cpp 7. replace all free() by Free() in: trainLinear.c predictLinear.c linear.cpp 8. Add #include <R.h> #include <Rinternals.h> #include <R_ext/Rdynload.h> in linear.cpp
1. Incorporate changes from LIBLINEAR versions 1.51 to 1.80: - Use set_print_string_function to set the print function - Add free_model_content and free_and_destroy_model functions (avoid memory problem if users declare a model variable) - Add check_probability_model (consistent with libsvm) - A new solver: coordinate descent for dual logistic regression - New optimization method for l1-regularized logistic regression - linear.cpp: * Use 1-norm stopping condition for l1-regularized solvers * newton_iter < l/10 replaced by newton_iter <= l/10 in l2r_lr_dual (for l < 10) 2. predict.LiblineaR function can return additional information: - probabilities (only for logistic regression models) - decision values