Research map / Probabilistic and scientific modeling / Causal discovery
Identifying dependent components from multi-domain linear mixtures
Danru Xu, Lauri Parkkonen, Sara Magliacane, Aapo Hyvarinen
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use MuDo-CoM instead of FastICA on correlated latents | FastICA · Simulated, ER-2 Gaussian SCM, n=10 | d-MCC: 0.783 -> 0.999 | Improved |
| use MuDo-CoM instead of QNDIAG on correlated latents | QNDIAG · Simulated, ER-2 Gaussian SCM, n=10 | d-MCC: 0.823 -> 0.999 | Improved |
| increase latent dependency degree from ER-0 to ER-3 | FastICA · Simulated, n=10 | d-MCC: 1.000 -> 0.745 | Hurt |
| increase latent dependency degree from ER-0 to ER-3 | MuDo-CoM · Simulated, n=10 | d-MCC: 0.994 -> 0.999 | No clear effect |
| use MuDo-CoM when latents are independent (ER-0) | FastICA and QNDIAG · Simulated, ER-0 | d-MCC: 1.000 vs 0.994 | Hurt |
| increase latent dimension n from 5 to 20 | PARAFAC · Simulated, ER-2 | d-MCC: 0.688 -> 0.319 | Hurt |
| increase latent dimension n from 5 to 20 | MuDo-CoM · Simulated, ER-2 | d-MCC: 0.999 -> 0.998 | No clear effect |
| increase sample size N from 200 to 10000 | MuDo-CoM · Simulated, n=10 | d-MCC: 0.900 -> 1.000 | Improved |
| increase number of domains T from 1n to 3n | MuDo-CoM · Simulated, n=10 | d-MCC: 0.998 -> 0.991 | No clear effect |
| increase observed dimension d from n to 10n | MuDo-CoM · Simulated, n=10 | d-MCC: 0.999 -> 0.999 | No clear effect |
| use exponential or Gumbel noise instead of Gaussian | MuDo-CoM · Simulated, n=10 | d-MCC: 0.999 -> 0.991 (Exp), 0.996 (Gumbel) | No clear effect |
| use nonlinear SCM instead of linear | MuDo-CoM · Simulated, n=10 | d-MCC: 0.999 -> 0.999 | No clear effect |
Closest papers by meaning
- Diverse Influence Component Analysis: A Geometric Approach to Nonlinear Mixture Identifiability (NeurIPS 2025 · similarity 0.87)
- Generalizing Nonlinear ICA Beyond Structural Sparsity (NeurIPS 2023 · similarity 0.86)
- On the Identifiability of Nonlinear ICA: Sparsity and Beyond (NeurIPS 2022 · similarity 0.86)
- Identification of Latent Confounders via Investigating the Tensor Ranks of the Nonlinear Observations (ICML 2025 · similarity 0.85)
- Probing the Latent Hierarchical Structure of Data via Diffusion Models (ICLR 2025 · similarity 0.83)
- Identifying General Mechanism Shifts in Linear Causal Representations (NeurIPS 2024 · similarity 0.83)
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.
