A Statistical Learning Analysis to Norm-Constrained Portfolio Optimization

在这个项目中，我们研究了一个投资组合优化方法的基础上的最小化的风险价值和CVaR，这已经获得了越来越多的流行在实践和理论。采用被称为统计学习方法ν-SVM的泛化理论的理论基础，我们实证地表明，范数约束版本的VaR和CVaR最小化比无约束版本实现了更好的样本外性能。

In this project, we investigate a portfolio optimization approach on the basis of the minimizations of VaR and CVaR, which have obtained a growing popularity both in practice and theory. Employing a theoretical underpinning known as the generalization theory for the ν-SVM, a statistical learning method, we empirically show that the norm-constrained versions of VaR and CVaR minimizations achieve better out-of-sample performance than the unconstrained versions.