Smooth Multi-Policy Causal Effect Estimation in Longitudinal Settings
Researchers have developed a new method called Policy-Encoded Q Network (PEQ-Net) to improve causal effect estimation in longitudinal settings. This approach allows for information sharing across different treatment policies, addressing a bias and variance issue found in traditional methods. PEQ-Net utilizes a shared policy encoder trained with kernel mean embeddings to reflect policy dissimilarities, leading to more stable and accurate results, especially when evaluating similar policies. AI
