Solving High-Order Portfolios via Successive Convex Approximation Algorithms

The first moment and second central moments of the portfolio return, a.k.a. mean variance, have been widely employed to assess expected profit risk portfolio. Investors pursue higher lower variance when designing portfolios. two can well describe distribution return it follows Gaussian distribution. However, real world assets is usually asymmetric heavy-tailed, which far from being a asymmetry heavy-tailedness are characterized by third fourth moments, i.e., skewness kurtosis, respectively. Higher kurtosis preferred reduce probability extreme losses. incorporating high-order in design very difficult due their non-convexity rapidly increasing computational cost with dimension. In this paper, we propose efficient convergence-provable algorithm framework based on successive convex approximation (SCA) solve efficiency proposed demonstrated numerical experiments.