Proximal gradient methods for learning: Difference between revisions

Revision as of 17:15, 10 December 2013

Revision as of 17:19, 10 December 2013

Line 116:

where <math>w_g</math> is the <math>g</math>th group.

In contrast to lasso, the derivation of the proximity operator for group lasso relies on the [[#Relevant_backround##Moreau_decomposition|Moreau decomposition]]. Here the proximity operator of the conjugate of the group lasso penalty becomes a projection onto the ball of a dual norm.<ref name=structSparse />

===Other group structures===