Tibshirani, Robert; Saunders, Michael; Rosset, Saharon; … - In: Journal of the Royal Statistical Society Series B 67 (2005) 1, pp. 91-108
The lasso penalizes a least squares regression by the sum of the absolute values ("L"<sub>1</sub>-norm) of the coefficients. The form of this penalty encourages sparse solutions (with many coefficients equal to 0). We propose the 'fused lasso', a generalization that is designed for problems with features...