Distributionally robust end-to-end portfolio construction

We propose an end-to-end distributionally robust system for portfolio construction that integrates the asset return prediction model with a optimization model. also show how to learn risk-tolerance parameter and degree of robustness directly from data. End-to-end systems have advantage in information can be communicated between decision layers during training, allowing parameters trained final task rather than solely predictive performance. However, existing are not able quantify correct impact risk on layer. Our proposed selection explicitly accounts risk. The layer chooses portfolios by solving minimax problem where distribution returns is assumed belong ambiguity set centered around nominal distribution. Using convex duality, we recast form allows efficient training system.