Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Efficient Uncertainty Quantification and Reduction for Over-Parameterized Neural Networks
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 |
|---|---|---|---|
| use PNC predictor instead of one base network | over-parameterized NN · Synthetic #1, d=2, n=128 | MSE: (6.68\pm 2.74)\times 10^{-4} -> (3.68\pm 1.84)\times 10^{-4} | Helped |
| use PNC predictor instead of one base network | over-parameterized NN · Synthetic #1, d=16, n=1024 | MSE: (3.16\pm 0.35)\times 10^{-3} -> (7.44\pm 0.99)\times 10^{-4} | Helped |
| PNC predictor vs deep ensemble with 2 networks (similar cost) | PNC predictor · Synthetic #1, d=2, n=1024 | MSE: (5.75\pm 2.28)\times 10^{-5} -> (8.92\pm 5.69)\times 10^{-6} | Helped |
| PNC predictor vs deep ensemble with 5 networks (2.5x runtime) | PNC predictor · Synthetic #1, d=8, n=1024 | MSE: (3.79\pm 1.01)\times 10^{-4} -> (2.02\pm 0.24)\times 10^{-4} | Helped |
| PNC-enhanced batching vs DropoutUQ intervals | PNC-enhanced batching · Synthetic #1 95% CI, d=16, n=1024 | coverage rate: 0.87 -> 0.95 | Helped |
| PNC-enhanced cheap bootstrap vs DropoutUQ intervals | PNC-enhanced cheap bootstrap · Synthetic #1 95% CI, d=8, n=256 | coverage rate: 0.88 -> 1.00 | Helped |
| PNC-enhanced batching vs cheap bootstrap interval width | PNC-enhanced batching · Synthetic #1 95% CI, d=16, n=1024 | interval width: 0.0550 -> 0.0388 | Helped |
| increase training size n from 128 to 1024 | PNC-enhanced batching · Synthetic #1 95% CI, d=2 | interval width: 0.0437 -> 0.0152 | Helped |
| increase training size n from 128 to 1024 | PNC predictor · Synthetic #1, d=2 | MSE: (3.68\pm 1.84)\times 10^{-4} -> (8.92\pm 5.69)\times 10^{-6} | Helped |
| use PNC-enhanced batching at small sample size | PNC-enhanced batching · Synthetic #1 90% CI, d=16, n=1024 | coverage rate: 0.92 | Mixed |
Closest papers by meaning
- Uncertainty Quantification with the Empirical Neural Tangent Kernel (NeurIPS 2025 · similarity 0.93)
- Epistemic Uncertainty Quantification For Pre-Trained Neural Networks (CVPR 2024 · similarity 0.90)
- Non-Asymptotic Uncertainty Quantification in High-Dimensional Learning (NeurIPS 2024 · similarity 0.89)
- Uncertainty Quantification for Deep Regression using Contextualised Normalizing Flows (NeurIPS 2025 · similarity 0.87)
- SURE: SUrvey REcipes for building reliable and robust deep networks (CVPR 2024 · similarity 0.87)
- Uncertainty Quantification for Physics-Informed Neural Networks with Extended Fiducial Inference (NeurIPS 2025 · similarity 0.87)
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.