Can we train machine learning methods to outperform the high-dimensional propensity score algorithm?

Related