Learning Causal Effect for High-dimensional Observation Data with Unmeasured Confounding
Published in Rejected in QE, Computer Science Department, Hong Kong Baptist University, 2023
Learning causal effects from high-dimensional observation data is critical for many realistic applications. The main challenge of causal effect estimation is the confounding problem, which includes both measured and unmeasured confounding, especially in high-dimensional data. In this survey, we have explored various frameworks, objectives, metrics, approaches, datasets, and packages for causal effect learning from observation data. We have also shared our preliminary work and future plans for learning causal effects from high-dimensional observation data. By addressing the confounding problem in causal effect estimation, we can develop better models and algorithms to uncover causal relationships and make more accurate predictions and decisions in real-world scenarios.