基于稀疏鲁棒M-投资选择模型的鲁棒Half算法

Robust half threshold algorithms based on the sparse robust M-Portfolios model

为得到鲁棒、稀疏的投资组合,提出稀疏鲁棒M-投资选择模型,并且基于L1/2正则化理论和Half阈值算法,构建鲁棒Half阈值算法求解稀疏鲁棒M-投资选择问题.数值实验表明,该算法不仅比Lasso算法收敛速度更快,而且在期望值固定的情况下得到的风险更小、更平稳.

The sparse and robust M-portfolio model is proposed to obtain robust and sparse portfolio,based on L1/2 regularization theory and the half threshold algorithm,the robust Half threshold algorithm is designed for numerically solving the sparse and robust M-portfolio problems.Numerical experiments show that this algorithm not only converges much faster than Lasso but also obtains a much less and much stabler risk obtained when the expectation value is fixed.