Re: [R] Efficient algorithm to get a solution path for ridge regression?
Thanks for your pointer. I looked into "Applied Regression Analysis" by Draper and Smith, and the ridge trace solution can be obtained efficiently by expressing it in canonical form. It could be coded without much difficulty, but I'm just wondering if there's any package which has already implemented this (or other) efficient way of returning a solution path for a sequence of regularization parameter? The function provided by MASS computes the solution individually for each value of the parameter. A quick look into the parcor and penalized packages seems to be promising, but I need to check the source code to see if they really implement an efficient algorithm as desired. Any input to this would be much appreciated. Thanks again. On 15-Apr-10, at 8:49 AM, Charles C. Berry wrote: On Wed, 14 Apr 2010, Kenneth Lo wrote: With the use of the LARS algorithm, a path of solutions corresponding to a sequence of the regularization parameter can be obtained for LASSO (or even the elastic net, a hybrid between LASSO and ridge) at the cost of one linear regression. In terms of computational speed LASSO seems to have beaten ridge regression, the solution of which needs to be computed individually, at the cost of one linear regression, for each regularization parameter. Is there any efficient method to compute a path of solutions for ridge regression corresponding to a sequence of the regularization parameter? Thanks. Yes. Check a textbook like Draper and Smith. Or Google for course notes. HTH, Chuck __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. Charles C. Berry(858) 534-2098 Dept of Family/Preventive Medicine E mailto:cbe...@tajo.ucsd.edu UC San Diego http://famprevmed.ucsd.edu/faculty/cberry/ La Jolla, San Diego 92093-0901 __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
Re: [R] Efficient algorithm to get a solution path for ridge regression?
On Wed, 14 Apr 2010, Kenneth Lo wrote: With the use of the LARS algorithm, a path of solutions corresponding to a sequence of the regularization parameter can be obtained for LASSO (or even the elastic net, a hybrid between LASSO and ridge) at the cost of one linear regression. In terms of computational speed LASSO seems to have beaten ridge regression, the solution of which needs to be computed individually, at the cost of one linear regression, for each regularization parameter. Is there any efficient method to compute a path of solutions for ridge regression corresponding to a sequence of the regularization parameter? Thanks. Yes. Check a textbook like Draper and Smith. Or Google for course notes. HTH, Chuck __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. Charles C. Berry(858) 534-2098 Dept of Family/Preventive Medicine E mailto:cbe...@tajo.ucsd.edu UC San Diego http://famprevmed.ucsd.edu/faculty/cberry/ La Jolla, San Diego 92093-0901 __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
[R] Efficient algorithm to get a solution path for ridge regression?
With the use of the LARS algorithm, a path of solutions corresponding to a sequence of the regularization parameter can be obtained for LASSO (or even the elastic net, a hybrid between LASSO and ridge) at the cost of one linear regression. In terms of computational speed LASSO seems to have beaten ridge regression, the solution of which needs to be computed individually, at the cost of one linear regression, for each regularization parameter. Is there any efficient method to compute a path of solutions for ridge regression corresponding to a sequence of the regularization parameter? Thanks. __ R-help@r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.