Research map / Label-efficient and robust learning / Recognition and long-tail
Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representations
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 |
|---|---|---|---|
| Increase regularization strength | Considered VAE methods except AnnealedVAE · Color-dSprites | total correlation of sampled representation | Helped |
| Increase regularization strength | Considered VAE methods · Color-dSprites | total correlation of mean representation | Hurt |
| Directly penalize off-diagonal covariance of the mean representation | DIP-VAE-I · Seven datasets | total correlation of mean representation | Helped |
| Choose the objective function rather than varying hyperparameters and random seeds | Unsupervised disentanglement methods · Seven datasets | disentanglement score variance explained: 37% ... 59% | Mixed |
| Select hyperparameters using unsupervised scores | Unsupervised disentanglement models · Shapes3D | rank correlation with disentanglement metrics | No clear effect |
| Transfer selected hyperparameters across datasets and metrics | Unsupervised disentanglement models · Seven datasets | probability of outperforming or equaling random model selection: 54.9% ... 59.3% | No clear effect |
| Increase disentanglement score | Learned representations · dSprites downstream classification | sample efficiency | No clear effect |
Closest papers by meaning
- Disentangling Factors of Variations Using Few Labels (ICLR 2020 · similarity 0.93)
- Theory and Evaluation Metrics for Learning Disentangled Representations (ICLR 2020 · similarity 0.93)
- On Disentangled Representations Learned from Correlated Data (ICML 2021 · similarity 0.93)
- Transferring disentangled representations: bridging the gap between synthetic and real images (NeurIPS 2024 · similarity 0.90)
- Disentangled Representation Learning in Non-Markovian Causal Systems (NeurIPS 2024 · similarity 0.90)
- Disentangled Representation Learning via Flow Matching (NeurIPS 2026 · similarity 0.89)
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.