The paper introduces a novel framework that integrates Ordinary Differential Equation (ODE) discovery methods into treatment effects inference. It aims to improve interpretability and performance by connecting these two fields. The authors propose the INSITE method, which consists of discovering population differential equations and fine-tuning them to obtain patient-specific differential equations. The method is evaluated on synthetic datasets generated from diverse pharmacological models, showcasing its effectiveness in predicting treatment effects over time. INSITE outperforms existing state-of-the-art methods in terms of counterfactual prediction error across various scenarios.
翻譯成其他語言
從原文內容
arxiv.org
深入探究