Code for causal isotonic calibration for heterogeneous treatment effects (appeared in ICML, 2023)
-
Updated
Nov 14, 2026 - Python
Code for causal isotonic calibration for heterogeneous treatment effects (appeared in ICML, 2023)
Leakage-aware LendingClub default-risk prediction with logit, elastic net, CART, bagging, random forests, gains and lift screening, and cross-fitted DML.
Aspect-level sentiment + Louvain taste communities on 675k Amazon reviews, tested leakage-free: cross-fitting cuts 'significant' community effects from 28.8% to 1.7%.
Double ML for time series data
Ensemble methods for learning features of heterogeneous treatment effects with valid downstream inference via repeated cross-fitting
Estimating causal treatment effects using Double Machine Learning (DoubleML), Interactive Regression Models, and Instrumental Variables with cross-fitting and orthogonalization
Code, Jupyter notebook and retained results for acquisition-level bridge monitoring on Z24, Route345 and openLAB: measurement-quality estimation, fault-aware classification, selective prediction, cross-fitting and instrumental verification. Companion repository for the ICAM3E 2026 study.
To associate your repository with the cross-fitting topic, visit your repo's landing page and select "manage topics."