DeCaFlow: A Deconfounding Causal Generative Model

Published in ArXiv Preprint, 2025

Consult the paper here

A practical Causal Generative Model that allows for estimating interventional distributions and counterfactuals,¡ even with unobserved confounders are present. DeCaFlow leverages variational inference and Causal Normalizing Flows to model a Confounded Structural Causal Models, that yields in unbiased estimations of causal queries that are identifiable through adjustment and Proximal Inference.

Cite as:

@article{almodovar2025decaflow,
  title={DeCaFlow: A Deconfounding Causal Generative Model},
  author={Almod{\'o}var, Alejandro and Javaloy, Adri{\'a}n and Parras, Juan and Zazo, Santiago and Valera, Isabel},
  journal={arXiv preprint arXiv:2503.15114},
  year={2025}
}
}