Utilizing meta-learning to recommend causal effect estimators
Estimation of causal effects using machine learning methods has become an active research field. Machine learning estimators as such are, however, primarily designed for prediction problems and thus cannot be used directly for causal inference. Therefore, new approaches for the estimation of causal parameters using machine learning emerged, such as meta-learners, specialized variants of tree-based models, and dedicated neural network architectures for estimating causal effects. The goal of this thesis is to utilize meta-learning techniques to recommend estimators for the estimation of heterogeneous treatment effects. More specifically, build a meta-model that learns the relationship between the characteristics of the task at hand and predicts the best meta-learner. This work will focus on the estimators implemented in the EconML library https://econml.azurewebsites.net.
Lõputöö kaitsmise aasta
Dr. Egor Kraev (Wise Plc) and Prof. Radwa El Shawi
Some background in causal inference or interest in learning it