|
Lasso and Dantzig are both methods for finding sparse solutions to linear regression. Which one works better for real data sets, say such as MRI or gene sequences? |
|
It depends and I think it might be hard to say anything in general. There is a recent paper Connections between the Dantzig selector and Lasso that gives some comparisons, and also discusses conditions under which Lasso and Dantzig actually turn out to be equivalent. |
|
As far I know, the Dantzig selector was introduced for theoretical reasons (basically the mathematicians -Tao- couldn't get proofs with the Lasso), whereas the Lasso is very much used in practice. I suspect that from a practical point of view you want to use the Lasso, although I can't really point to a specific reason.
This answer is marked "community wiki".
|
