Research map / Probabilistic and scientific modeling / Variational and generative sampling
Disentangling Learning Representations with Density Estimation
Eric Yeats, Frank Y Liu, Hai Li
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 |
|---|---|---|---|
| Add positive disentanglement regularization (λ > 0) (ablation) | GCAE · dSprites | MIG: ~0.03 -> ~0.35 | Improved |
| Add positive disentanglement regularization (λ > 0) (ablation) | GCAE · dSprites | FactorScore: ~0.47 -> ~0.6 | Improved |
| Add positive disentanglement regularization (λ > 0) (ablation) | GCAE · dSprites | DCI Disentanglement: ~0.03 -> ~0.45 | Improved |
| Add positive disentanglement regularization (λ > 0) (ablation) | GCAE · dSprites | SAP: ~0.08 -> ~0.37 | Improved |
| Use σ={0.2, 0.3} rather than σ=0.1 (ablation) | GCAE · dSprites | disentanglement scores | Improved |
| Use σ={0.2, 0.3} rather than σ=0.1 (ablation) | GCAE · Beamsynthesis | disentanglement scores | Improved |
| Increase latent dimension from m=10 to m=20 (ablation) | GCAE · dSprites | average disentanglement scores | No clear effect |
| Increase latent dimension from m=10 to m=20 (ablation) | FactorVAE · dSprites | disentanglement scores | Hurt |
| Use Excess Entropy Power loss (LEEP) instead of direct gradient descent on LSI (ablation) | GCAE · disentanglement experiments | MIG stability | Improved |
| Increase λ and σ (ablation) | GCAE · Beamsynthesis and dSprites | MSE reconstruction error | Hurt |
| Increase λ and σ (ablation) | GCAE · Beamsynthesis and dSprites | LSI | Improved |
| Use GCAE rather than FactorVAE | disentanglement methods · dSprites, m=20 versus m=10 | robustness to latent dimension | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- FactorVAE — FactorVAE: A Probabilistic Dynamic Factor Model Based on Variational Autoencoder for Predicting Cross-Sectional Stock Returns (AAAI 2022) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Isolating Sources of Disentanglement in Variational Autoencoders (NeurIPS 2018 · similarity 0.89)
- Disentangling Disentanglement in Variational Autoencoders (ICML 2019 · similarity 0.89)
- Orthogonality-Enforced Latent Space in Autoencoders: An Approach to Learning Disentangled Representations (ICML 2023 · similarity 0.89)
- Disentangling Multiple Features in Video Sequences using Gaussian Processes in Variational Autoencoders (ECCV 2020 · similarity 0.88)
- $\alpha$TC-VAE: On the relationship between Disentanglement and Diversity (ICLR 2024 · similarity 0.86)
- Sequential Disentanglement by Extracting Static Information From A Single Sequence Element (ICML 2024 · similarity 0.85)
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.
