Bayesian Portfolio Regularization

关于投资组合模型表现的程式化事实表明，与基于更简单和理论上较差的策略的方法相比，理论上最优投资组合策略的经验对应物表现出非常差的表现。提出的研究项目的目标是开发一类新的贝叶斯正则化策略，以实现稳健和高性能的投资组合模型。贝叶斯方法直接应用于组合权重的估计，为大范围的不同正则化策略以及模型选择和评估提供了统一的框架。除了理论贡献外，该项目还将提供关于针对各种资产和市场提出的新策略质量的经验证据。将特别注意在高维环境下的估计策略。

Stylized facts on the performance of portfolio models indicate that the empirical counterparts of theoretically optimal portfolio strategies show a very poor performance compared to approaches based on simpler and theoretically inferior strategies. The goal of the proposed research project is to develop a new class of Bayesian regularization strategies to achieve robust and high-performing portfolio models. The Bayesian approach applied directly to the estimation of the portfolio weights provides a unifying framework for a large range of different regularization strategies as well as for model selection and evaluation.Besides the theoretical contributions the project will also provide empirical evidence on the quality of the proposed new strategies for a variety of assets and markets. Special attention will be given to estimation strategies in high-dimensional settings.