AI Paper Atlas

Research map / Learning theory and algorithms / Matrix and tensor methods

Disentangling by Subspace Diffusion

David Pfau, Irina Higgins, Alex Botev, Sébastien Racanière

NeurIPS 2020 · Learning theory and algorithms › Matrix and tensor methods · approach: High-dimensional estimation theory · 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
Apply a random orthogonal rotation to the true latent state (ablation)GEOMANCER · Stanford Bunnyaverage angle: 0.024 -> 0.024No clear effect
Randomly scale individual latent dimensions (ablation)GEOMANCER · Stanford Bunnyaverage angle: 0.024 -> 0.72±0.68Hurt
Apply a random linear transformation to the true latent state (ablation)GEOMANCER · Stanford Bunnyaverage angle: 0.024 -> 1.42±0.09Hurt
Use a Laplacian Eigenmaps embedding instead of the true latent state (ablation)GEOMANCER · Stanford Bunnyaverage angle: 0.024 -> 0.37; Chance 1.26±0.23Mixed
Use pixels as input instead of the true latent state (ablation)GEOMANCER · Stanford Bunnyaverage angle: 1.25; Chance 1.26±0.23No clear effect
Use β-VAE latent vectors instead of the true latent state (ablation)GEOMANCER · Stanford Bunnyaverage angle: 1.30±0.02; Chance 1.26±0.23No clear effect
Apply a random orthogonal rotation to the true latent state (ablation)GEOMANCER · Stanford Dragonaverage angle: 0.023 -> 0.023No clear effect
Randomly scale individual latent dimensions (ablation)GEOMANCER · Stanford Dragonaverage angle: 0.023 -> 0.96±0.70Hurt
Apply a random linear transformation to the true latent state (ablation)GEOMANCER · Stanford Dragonaverage angle: 0.023 -> 1.26±0.37Hurt
Use a Laplacian Eigenmaps embedding instead of the true latent state (ablation)GEOMANCER · Stanford Dragonaverage angle: 0.023 -> 0.32; Chance 1.26±0.23Mixed
Use pixels as input instead of the true latent state (ablation)GEOMANCER · Stanford Dragonaverage angle: 1.26; Chance 1.26±0.23No clear effect
Use β-VAE latent vectors instead of the true latent state (ablation)GEOMANCER · Stanford Dragonaverage angle: 1.15±0.09; Chance 1.26±0.23No clear effect

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