Research map / Probabilistic and scientific modeling / Variational and generative sampling
Pseudo-Non-Linear Data Augmentation: A Constrained Energy Minimization Viewpoint
Pingbang Hu, Mahito Sugiyama
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 |
|---|---|---|---|
| augment training set with PNL samples (20% extra) | MLP · Connectionist Bench | test accuracy (%): 88.10 -> 93.81 | Improved |
| augment training set with PNL samples | MLP · Wine Quality | test accuracy (%): 55.00 -> 59.03 | Improved |
| augment training set with PNL samples | ResNet-18 · CIFAR-10 | test accuracy (%): 88.57 -> 88.07 | Hurt |
| augment training set with PNL samples | MLP · MNIST | test accuracy (%): 97.98 -> 97.91 | No clear effect |
| augment training set with PNL samples | MLP · Taiwanese Bankruptcy | test accuracy (%): 96.54 -> 96.53 | No clear effect |
| augment training set with PNL samples | M5 · Speech Commands | test accuracy (%): 84.48 -> 84.35 | No clear effect |
| use PNL instead of autoencoder augmentation | MLP · Connectionist Bench | test accuracy (%): 82.86 -> 93.81 | Improved |
| use PNL instead of mixup | MLP · MNIST | test accuracy (%): 96.45 -> 97.91 | Improved |
| PNL lowers variance across bootstrap tests | MLP · Connectionist Bench | accuracy standard deviation: 8.58% -> 4.54% | Improved |
| increase PNL augmentation size from 25% to 100% of original (ablation) | MLP · MNIST | test accuracy (%): 98.01 -> 97.55 | Hurt |
| use k=1 nearest neighbor for the local data sub-manifold (ablation) | PNL · MNIST (qualitative) | sample diversity | Hurt |
| apply direct kernel density fitting in ambient space without dimension reduction (ablation) | PNL · MNIST (1000 samples) | augmentation quality | Hurt |
Closest papers by meaning
- Understanding Augmentation-based Self-Supervised Representation Learning via RKHS Approximation and Regression (ICLR 2024 · similarity 0.86)
- Tradeoffs in Data Augmentation: An Empirical Study (ICLR 2021 · similarity 0.86)
- Augmentation with Projection: Towards an Effective and Efficient Data Augmentation Paradigm for Distillation (ICLR 2023 · similarity 0.85)
- Automatic Data Augmentation via Invariance-Constrained Learning (ICML 2023 · similarity 0.85)
- How Data Augmentation Shapes Neural Representations (NeurIPS 2026 · similarity 0.83)
- UtilGen: Utility-Centric Generative Data Augmentation with Dual-Level Task Adaptation (NeurIPS 2025 · similarity 0.83)
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.
