Research map / Label-efficient and robust learning / Recognition and long-tail
Leveraging sparse and shared feature activations for disentangled representation learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| add sparse sufficiency and feature sharing regularizers | representation learning model · DSprites, 3DShapes, SmallNorb, and Cars | DCI disentanglement: 16.6 -> 69.9; 44.4 -> 87.7; 16.5 -> 55.8; 60.5 -> 92.3 | Helped |
| add sparse sufficiency and feature sharing regularizers | representation learning model · PACS | OOD accuracy (average): 85.7 -> 87.5 | Helped |
| add sparse sufficiency and feature sharing regularizers | representation learning model · VLCS | OOD accuracy (average): 77.4 -> 78.4 | Helped |
| add sparse sufficiency and feature sharing regularizers | representation learning model · OfficeHome | OOD accuracy (average): 67.5 -> 70.9 | Helped |
| add sparse sufficiency and feature sharing regularizers | representation learning model · Waterbirds | OOD accuracy (average): 81.3 -> 90.5 | Helped |
| add sparse sufficiency and feature sharing regularizers | Camelyon17 model · Camelyon17 test OOD | accuracy: 70.3 -> 74.1 ±0.2 | Helped |
| remove feature sharing regularizer (ablation) | DSprites representation · DSprites | DCI disentanglement: 98.8 -> 71.9 | Hurt |
| increase feature sharing weight beta to 0.4 (ablation) | DSprites representation · DSprites | DCI disentanglement: 98.8 -> 30.5 | Hurt |
| add sparsity regularizer only (alpha=1e-2, beta=0) (ablation) | DSprites representation · DSprites | DCI disentanglement: 27.8 -> 71.9 | Helped |
| add feature sharing with beta=0.2 while retaining sparsity (ablation) | DSprites representation · DSprites | DCI disentanglement: 71.9 -> 98.8 | Helped |
| add sparse sufficiency and feature sharing regularizers | AbstractDSprites representation · AbstractDSprites compositional tasks, |S|=3 | accuracy: 72.6 -> 83.0 | Helped |
Closest papers by meaning
- Disentangling Representations through Multi-task Learning (ICLR 2025 · similarity 0.90)
- Disentangled Representation Learning via Flow Matching (NeurIPS 2026 · similarity 0.90)
- Graph-based Unsupervised Disentangled Representation Learning via Multimodal Large Language Models (NeurIPS 2024 · similarity 0.89)
- Synergy Between Sufficient Changes and Sparse Mixing Procedure for Disentangled Representation Learning (ICLR 2025 · similarity 0.89)
- An Information Criterion for Controlled Disentanglement of Multimodal Data (ICLR 2025 · similarity 0.89)
- Transferring disentangled representations: bridging the gap between synthetic and real images (NeurIPS 2024 · similarity 0.88)
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.