-
- Description:
DAL is an efficient and flexibible MATLAB toolbox for solving the following optimization problem:
minimize f(Ax) + lambda*c(x)
where A (m x n) is a design matrix, f is a loss function, and c is a measure of sparsity.
DAL can handle your favorite (convex, smooth) loss functions (squared loss, logistic loss, etc).
DAL can handle A (and its transpose) provided as function handles.
DAL can handle several "sparsity" measures in an unified way. Currently L1, grouped L1, and trace norm (testing, requires PROPACK) measures are supported.
DAL is efficient when m<<n (m: #samples, n: #unknowns) or the matrix A is poorly conditioned.
DAL is fast when the solution is sparse but the matrix A can be dense.
DAL is written in MATLAB.
- Changes to previous version:
- Supports weighted lasso (dalsqal1.m, dallral1.m)
- Supports weighted squared loss (dalwl1.m)
- Bug fixes (group lasso and elastic-net-regularized logistic regression)
- BibTeX Entry: Download
- Corresponding Paper BibTeX Entry: Download
- Supported Operating Systems: Agnostic
- Data Formats: Agnostic
- Tags: Optimization, Trace Norm, Group Lasso, Lasso, Sparse Learning, L1 Regularization, Logistic Regression
- Archive: download here
Other available revisons
-
Version Changelog Date 1.1 - Supports weighted lasso (dalsqal1.m, dallral1.m)
- Supports weighted squared loss (dalwl1.m)
- Bug fixes (group lasso and elastic-net-regularized logistic regression)
February 18, 2014, 19:07:06 1.05 - 35% faster group lasso.
- Sparse connectivity inference example added (s_test_hsgl.m).
- Non-negative lasso (thanks to Shigeyuki Oba).
- Uses Mark Tygert's pca.m for SVD (PROPACK is not required anymore).
May 3, 2011, 07:00:43 1.01 - Logistic loss: : dallrl1.m, dallrgl.m, dallrds.m
- Unequal-sized blocks supported in Group lasso regularization
- eta: initial eta=0.01/lambda
- dallrds.m: trace-norm regularized logistic regression (requires PROPACK)
December 14, 2009, 09:43:50 0.97 Initial Announcement on mloss.org.
April 13, 2009, 09:39:59
Comments
No one has posted any comments yet. Perhaps you'd like to be the first?
Leave a comment
You must be logged in to post comments.