Theory for ℓ1/ℓ2-penalty procedures
We study four procedures for regression models with group structure in the parameter vector. The first two are for models with univariate response variable. They are the so-called group Lasso (see Chapter 4), and the smoothed group Lasso for the high-dimensional additive model (see Chapter 5). We also discuss multivariate extensions, namely for the linear model with time-varying coefficients, for multivariate regression, and multitask learning.
KeywordsRegularization Parameter Empirical Process Error Loss Truncation Level Group Lasso
Unable to display preview. Download preview PDF.