Research map / Probabilistic and scientific modeling / Variational and generative sampling
A Bayesian Nonparametric Framework For Learning Disentangled Representations
Vaishnavi S Patil, Siddhi Patil, Matthew Evanusa, Amit Kumar Kundu, Cornelia Fermuller, Joseph JaJa
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 |
|---|---|---|---|
| use nested instead of truncated (k=10) variational family (ablation) | Bayes-QLAE · Ablation (Table 4) | InfoC: 0.40 ± .03 -> 0.51 ± .03 | Improved |
| use nested instead of truncated (k=10) variational family (ablation) | Bayes-QLAE · Ablation (Table 4) | InfoM: 0.54 ± .04 -> 0.58 ± .04 | Improved |
| use nested instead of truncated (k=50) variational family (ablation) | Bayes-QLAE · Ablation (Table 4) | InfoM: 0.51 ± .06 -> 0.58 ± .04 | Improved |
| use nested instead of truncated (k=50) variational family (ablation) | Bayes-QLAE · Ablation (Table 4) | InfoC: 0.48 ± .05 -> 0.51 ± .03 | Improved |
| use structured instead of mean-field variational family (ablation) | Bayes-QLAE · Ablation (Table 4) | InfoM: 0.49 ± .04 -> 0.58 ± .04 | Improved |
| use structured instead of mean-field variational family (ablation) | Bayes-QLAE · Ablation (Table 4) | InfoC: 0.49 ± .04 -> 0.51 ± .03 | Improved |
| use stochastic instead of deterministic quantization (ablation) | Bayes-QLAE · Ablation (Table 4) | InfoM: 0.52 ± .02 -> 0.58 ± .04 | Improved |
| use stochastic instead of deterministic quantization (ablation) | Bayes-QLAE · Ablation (Table 4) | InfoC: 0.43 ± .02 -> 0.51 ± .03 | Improved |
| use stochastic instead of deterministic quantization (ablation) | Bayes-QLAE · Ablation (Table 4) | InfoE: 0.71 ± .02 -> 0.71 ± .04 | No clear effect |
| use structured instead of mean-field variational family (ablation) | Bayes-QLAE · Ablation (Table 4) | InfoE: 0.76 ± .04 -> 0.71 ± .04 | Hurt |
| replace QLAE latent quantization with Bayes-QLAE nonparametric prior | QLAE · 3DShapes | InfoC: 0.49 ± .01 -> 0.61 ± .02 | Improved |
| replace QLAE latent quantization with Bayes-QLAE nonparametric prior | QLAE · MPI3D | InfoC: 0.43 ± .02 -> 0.56 ± .03 | 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.
- Tripod — Tripod: Three Complementary Inductive Biases for Disentangled Representation Learning (ICML 2024) · better on 5 of 6 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
- Variational Inference of Disentangled Latent Concepts from Unlabeled Observations (ICLR 2018 · similarity 0.91)
- Disentanglement via Latent Quantization (NeurIPS 2023 · similarity 0.91)
- $\alpha$TC-VAE: On the relationship between Disentanglement and Diversity (ICLR 2024 · similarity 0.90)
- Synergy Between Sufficient Changes and Sparse Mixing Procedure for Disentangled Representation Learning (ICLR 2025 · similarity 0.90)
- Disentangled Representation Learning via Flow Matching (NeurIPS 2026 · similarity 0.89)
- Disentangled Representation Learning via Modular Compositional Bias (NeurIPS 2025 · 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.
