Counterfactual Identifiability of Bijective Causal Models

2023/02/04 by Arash Nasr-Esfahany, Mohammad Alizadeh, Nasr-Esfahany, Arash +3 · 8 citations
Computer Science · #Error Correcting Code Techniques #FOS: Computer and information sciences #Generative Adversarial Networks and Image Synthesis #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Music and Audio Processing

paper · pdf · doi:10.48550/arxiv.2302.02228

openalex publication_date 2023/02/04 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

We study counterfactual identifiability in causal models with bijective generation mechanisms (BGM), a class that generalizes several widely-used causal models in the literature. We establish their counterfactual identifiability for three common causal structures with unobserved confounding, and propose a practical learning method that casts learning a BGM as structured generative modeling. Learned BGMs enable efficient counterfactual estimation and can be obtained using a variety of deep conditional generative models. We evaluate our techniques in a visual task and demonstrate its application in a real-world video streaming simulation task.

Cited by

• DoFlow: Causal Generative Flows for Interventional and Counterfactual Time-Series Prediction
• Counterfactual Identifiability via Dynamic Optimal Transport
• Counterfactually Fair Conformal Prediction
• Wasserstein Distributionally Robust Optimization Through the Lens of Structural Causal Models and Individual Fairness
• Causal-Adapter: Taming Text-to-Image Diffusion for Faithful Counterfactual Generation
• What is a good matching of probability measures? A counterfactual lens on transport maps
• Partially Functional Dynamic Backdoor Diffusion-based Causal Model
• Flow IV: Counterfactual Inference In Nonseparable Outcome Models Using Instrumental Variables

Related

• Counterfactual (Non-)identifiability of Learned Structural Causal Models
• High Fidelity Image Counterfactuals with Probabilistic Causal Models
• Unsupervised Structural-Counterfactual Generation under Domain Shift
• The Causal Round Trip: Generating Authentic Counterfactuals by Eliminating Information Loss
• Learning Generalized Gumbel-max Causal Mechanisms