Álvaro Parafita
@aparafita.bsky.social
Senior Researcher at Barcelona Supercomputing Center | PhD in Causal Estimation with estimand-agnostic frameworks, working on Machine Learning Explainability
Github: @aparafita
Github: @aparafita
The goal is to bridge Causal Inference with Explainable AI, providing estimand-agnostic strategies to facilitate the use of these techniques for non-experts.
See you at #NeurIPS25!
See you at #NeurIPS25!
September 25, 2025 at 3:13 PM
The goal is to bridge Causal Inference with Explainable AI, providing estimand-agnostic strategies to facilitate the use of these techniques for non-experts.
See you at #NeurIPS25!
See you at #NeurIPS25!
We present a practical extension of do-SHAP that removes reliance on estimands, enabling causal feature attribution on complex graphs.
We additionally introduce the Frontier-Reducibility Algorithm (FRA), which significantly accelerates do-SHAP estimation with negligible computational overhead.
We additionally introduce the Frontier-Reducibility Algorithm (FRA), which significantly accelerates do-SHAP estimation with negligible computational overhead.
September 25, 2025 at 3:13 PM
We present a practical extension of do-SHAP that removes reliance on estimands, enabling causal feature attribution on complex graphs.
We additionally introduce the Frontier-Reducibility Algorithm (FRA), which significantly accelerates do-SHAP estimation with negligible computational overhead.
We additionally introduce the Frontier-Reducibility Algorithm (FRA), which significantly accelerates do-SHAP estimation with negligible computational overhead.