Properties and iterative methods for the lasso and its variants (Q741471): Difference between revisions

The author introduces several new iterative algorithms based on the proximal map and proves the convergence of the generated sequences. The results are applied to approximate a solution of the minizimization problem \[ \min_{x}{1 \over 2} \|Ax-b\|_{2}^{2} {\;subject\;to\;}\|x\|_{1}\leq t, \] known as least absolute shrinkage operator (or lasso) and first introduced in [\textit{R. Tibshirani}, J. R. Stat. Soc., Ser. B 58, No. 1, 267--288 (1996; Zbl 0850.62538)]. Convergence theorems for solving two variants of lasso are also proved.