Research map / Probabilistic and scientific modeling / Variational and generative sampling
NashAE: Disentangling Representations through Adversarial Covariance Minimization
Eric Yeats, Frank Liu, David Womble, 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 adversarial covariance loss (lambda 0 -> 0.2) (ablation) | NashAE · CelebA | TAD: 0.235 -> 0.543 | Improved |
| raise lambda from 0.2 to 0.8 (ablation) | NashAE · CelebA | TAD: 0.543 -> 0.474 | Hurt |
| add adversarial covariance loss (lambda 0 -> 0.002) (ablation) | NashAE · dSprites | beta-VAE metric (%): 91.41% -> 96.57% | Improved |
| add adversarial covariance loss (lambda 0 -> 0.2) (ablation) | NashAE · Beamsynthesis m=32 | abs. latent-count error: 28.75 -> 1 | Improved |
| add adversarial covariance loss (lambda 0 -> 0.3) (ablation) | NashAE · Beamsynthesis m=8 | abs. latent-count error: 5.75 -> 0 | Improved |
| raise lambda from 0.2 to 0.3 at m=32 (ablation) | NashAE · Beamsynthesis m=32 | abs. latent-count error: 1 -> 0.5 | Improved |
| raise lambda from 0 to 0.2 (ablation) | NashAE · CelebA latent fixed point | average R^2 of latent predictions: 0.52 -> 0.04 | Improved |
| raise lambda from 0 to 0.2 (ablation) | NashAE · CelebA latent fixed point | number of learned latents: 32 -> 22 | Improved |
| raise beta from 1 to 100 (ablation) | β-VAE · CelebA | TAD: 0.158 -> 0.351 | Improved |
| raise beta from 100 to 250 (ablation) | β-VAE · CelebA | TAD: 0.351 -> 0.307 | Hurt |
| raise beta from 1 to 8 on dSprites (ablation) | β-VAE · dSprites | beta-VAE metric (%): 84.63% -> 96.21% | Improved |
| raise beta from 2 to 4 (ablation) | β-TCVAE · dSprites | beta-VAE metric (%): 95.01% -> 93.95% | Hurt |
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 β=15 — FactorVAE: A Probabilistic Dynamic Factor Model Based on Variational Autoencoder for Predicting Cross-Sectional Stock Returns (AAAI 2022) · better on 7 of 8 shared results
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
- Disentangling Factors of Variation with Cycle-Consistent Variational Auto-Encoders (ECCV 2018 · similarity 0.91)
- Demystifying Inter-Class Disentanglement (ICLR 2020 · similarity 0.91)
- Disentangling Disentanglement in Variational Autoencoders (ICML 2019 · similarity 0.90)
- Learning Disentangled Representations with Latent Variation Predictability (ECCV 2020 · similarity 0.90)
- Disentangling by Factorising (ICML 2018 · similarity 0.90)
- Isolating Sources of Disentanglement in Variational Autoencoders (NeurIPS 2018 · 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.
