Descent-Based Calibrated Optimal Direct Estimation
Description
Algorithms for solving a self-calibrated l1-regularized quadratic programming problem without parameter tuning. The algorithm, called DECODE, can handle high-dimensional data without cross-validation. It is found useful in high dimensional portfolio selection (see Pun (2018) ) and large precision matrix estimation and sparse linear discriminant analysis (see Pun and Hadimaja (2019) ).