Research map / Probabilistic and scientific modeling / Causal discovery
Toward Causal Generative Modeling: From Representation to Generation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
A doctoral-dissertation summary describing the author's line of work on causal generative modeling: causal VAEs, independent-causal-mechanism disentanglement, and diffusion-based counterfactual image generation.
Problem
Generalization of deep representation learning under distribution shift, and how to recover disentangled causal factors and generate counterfactuals with high fidelity under weak supervision.
Approach
Combines structural causal models with generative models: causal priors and nonlinear SCMs in VAEs (SCM-VAE), autoregressive normalizing flows with a causally factorized prior for independent causal mechanisms (ICM-VAE), and a diffusion model conditioned on causal representations with classifier-free-guidance-style weak supervision (CausalDiffAE).
Key results
- The paper reports empirically that SCM-VAE learns representations closer to ground-truth causal variables, that ICM-VAE achieves high causal disentanglement and intervention robustness, and that CausalDiffAE enables high-fidelity counterfactual generation (no specific numbers given in this summary).
Limitations
- Summary gives no quantitative results or experimental details for the cited works.
- Future directions are described only at a high level.
Experimental setup
Builds on: SCM-VAE ICM-VAE CausalDiffAE classifier-free guidance independent component analysis
Closest papers by meaning
- Disentangled Representation Learning in Non-Markovian Causal Systems (NeurIPS 2024 · similarity 0.90)
- From Causal to Concept-Based Representation Learning (NeurIPS 2024 · similarity 0.88)
- CaRiNG: Learning Temporal Causal Representation under Non-Invertible Generation Process (ICML 2024 · similarity 0.88)
- A Fixed-Point Approach for Causal Generative Modeling (ICML 2024 · similarity 0.88)
- Learning Linear Causal Representations from General Environments: Identifiability and Intrinsic Ambiguity (NeurIPS 2024 · similarity 0.87)
- Counterfactual Generative Modeling with Variational Causal Inference (ICLR 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.