Research map / Label-efficient and robust learning / Noisy labels
Unified Robust Semi-Supervised Variational Autoencoder
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 uncertainty prior on mixture parameters (ablation) | URSVAE · CIFAR-10 / CIFAR-100 (sym. label noise, outliers) | accuracy: e.g. 95.7 vs 91.7 at 50/10 on CIFAR-10 | Helped |
| replace beta-divergence with KL (remove robust divergence) (ablation) | URSVAE · CIFAR-10 80% label noise, 10% outliers | accuracy: 92.6 without vs 93.3 with | Mixed |
| remove noise transition model (ablation) | URSVAE · CIFAR-10 sym. label noise (50% and 80%, 10% outliers) | accuracy: 86.5 vs 95.7 at 50/10; 90.7 vs 93.3 at 80/10 | Helped |
| add robust beta-divergence (tuned beta) over KL (ablation) | URSVAE · CIFAR-10 / CIFAR-100 | accuracy: about 2.2% on average per text | Helped |
| full URSVAE vs Dividemix and other baselines | semi-supervised noisy-label classification · CIFAR-10, CIFAR-100, Clothing1M, WebVision | accuracy: e.g. Clothing1M 79.65 vs 73.16 (Dividemix) | Helped |
Closest papers by meaning
- SeNM-VAE: Semi-Supervised Noise Modeling with Hierarchical Variational Autoencoder (CVPR 2024 · similarity 0.87)
- Denoising Criterion for Variational Auto-Encoding Framework (AAAI 2017 · similarity 0.86)
- Semi-Supervised Deep Regression with Uncertainty Consistency and Variational Model Ensembling via Bayesian Neural Networks (AAAI 2023 · similarity 0.86)
- Trading off Image Quality for Robustness is not Necessary with Regularized Deterministic Autoencoders (NeurIPS 2022 · similarity 0.86)
- SURE: SUrvey REcipes for building reliable and robust deep networks (CVPR 2024 · similarity 0.84)
- $t^3$-Variational Autoencoder: Learning Heavy-tailed Data with Student's t and Power Divergence (ICLR 2024 · similarity 0.84)
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.