Model-Free Risk-Neutral Density with Constrained Machine Learning:
I’m currently working on the estimation of predictive information for stocks from their options, together with my supervisors. We start from the Breeden and Litzenberger (1978) theorem and enhence the recovery of risk-neutral density (RND) by using machine learning with constraints. The proposed constrained-machine-learning-empowered RND-recovery method generates improved tail risk and moment estimations.
