AI Paper Atlas

Research map / Probabilistic and scientific modeling / Causal discovery

Learning Linear Causal Representations from Interventions under General Nonlinear Mixing

Simon Buchholz, Goutham Rajendran, Elan Rosenfeld, Bryon Aragam, Bernhard Schölkopf, Pradeep Kumar Ravikumar

NeurIPS 2023 · Probabilistic and scientific modeling › Causal discovery · approach: Causal representation identifiability · 4 citations (OpenAlex)

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 changedWhereMeasuredEffect
use nonlinear MLP encoder h instead of linear h on linear mixing (ablation)Contrastive · linear f (d=5, d'=10)MCC: 0.90 -> 0.05Hurt
nonlinear contrastive vs linear baseline for graph recovery on linear mixingContrastive · linear f (d=5, d'=10)SHD: 7.0 -> 4.6Improved
contrastive linear vs linear baseline for graph recovery on linear mixingContrastive Linear · linear f (d=5, d'=10)AUROC: 0.64 -> 0.80Improved
replace linear disentanglement baseline with nonlinear contrastive methodContrastive · nonlinear f, ER(5,2), d'=20MCC: 0.32 -> 0.97Improved
replace linear baseline with contrastive method on graph recoveryContrastive · nonlinear f, ER(10,2), d'=20SHD: 29.6 -> 3.6Improved
replace VAE with contrastive method for latent recoveryContrastive · nonlinear f, ER(10,2), d'=100R2: 0.78 -> 0.97Improved
increase observed dimension d' from 20 to 100 (ablation)Contrastive · nonlinear f, ER(5,2)MCC: 0.97 -> 0.99Improved
increase observed dimension d' from 20 to 100 (ablation)Linear baseline · nonlinear f, ER(5,2)MCC: 0.32 -> 0.18Hurt
increase latent dimension d from 5 to 10 (ablation)Contrastive · nonlinear f, d'=20SHD: 1.8 -> 3.6Hurt
replace VAE with contrastive learning on image dataContrastive Learning · image data, 2 ballsMCC: 0.19 -> 0.87Improved
replace VAE with contrastive learning on image dataContrastive Learning · image data, 5 ballsSHD: 18.6 -> 2.0Improved
increase number of balls from 5 to 10 (ablation)Contrastive Learning · image dataSHD: 2.0 -> 11.0Hurt

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy