hqreg
Regularization Paths for Lasso or Elastic-Net Penalized Huber Loss Regression and Quantile Regression
Efficient algorithms for fitting regularization paths for lasso or elastic-net penalized regression models with Huber loss, quantile loss or squared loss.
Regularization Paths for Huber Loss Regression and Quantile Regression Penalized by Lasso or Elastic-Net
To install:
- the released version from CRAN:
install.packages("hqreg") - the latest version (require
devtools):install_github("CY-dev/hqreg")
To report bugs：
- send email to: Congrui Yi [email protected]
News
1.0 * Added support for L2 penalization(alpha=0); * hqreg: changed argument "loss" to "method" and an option "squared" to "ls".
1.1 * loss.hqreg.R: corrected a mistake in function "qloss".
1.2 * cv.hqreg: added two new arguments, "fold.id" for predefined cross-validation folds, and "type.measure" for inclusion of general error measures "mse" and "mae". * loss.hqreg: modified definition depending on "type.measure".
1.3 * cv.hqreg: enabled parallel computing for cross-validation with a new argument "ncores" * hqreg: set the default value of "gamma" for Huber loss to be IQR(y)/10 * plot.hqreg, plot.cv.hqreg: changed argument "log.x" to "log.l" * removed computational redundancy and reduced memory allocation in source code * corrected typos in documentation
1.4 * new function "hqreg_raw" for fitting models on raw data without preprocessing * hqreg_raw contains a boolean argument "intercept" which can be set to FALSE to allow a "no-intercept" fit * hqreg: no longer support preprocess = "none" since the package now has a specific function hqreg_raw for that situation * modified plot, predict function accordingly for the situation without an intercept * cv.hqreg: added a new argument FUN that take values from either "hqreg" or "hqreg_raw" to include cross validation functionality for hqreg_raw * robustified the optimization algorithm for extreme cases in quantile regression when tau is close to 0 or 1