Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Reliable training and estimation of variance networks
Nicki Skafte, Martin Jørgensen, Søren Hauberg
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 the full Combined variance-estimation method instead of standard NN variance training | UCI regression models · UCI regression datasets | test log-likelihood | Improved |
| Use the full Combined method instead of standard VAE variance training | VAE · MNIST, FashionMNIST, CIFAR10, SVHN | test log-likelihood | Improved |
| Add locality sampling (ablation) | Combined variance network · Four UCI regression datasets | test log-likelihood | Improved |
| Add an inverse-Gamma variance distribution (ablation) | Combined variance network · Four UCI regression datasets | test log-likelihood | Improved |
| Add mean-variance split training (ablation) | Combined variance network · Four UCI regression datasets | test log-likelihood | Improved |
| Add the extrapolating variance architecture (ablation) | Combined variance network · Four UCI regression datasets | test log-likelihood | Improved |
| Combine locality sampling, split training, inverse-Gamma variance, and extrapolation (ablation) | Variance network · Four UCI regression datasets | test log-likelihood | Improved |
| Use locality sampling rather than ordinary mini-batches (ablation) | Variance network · Gradient experiments | gradient sparsity and stability | Improved |
| Use Combined uncertainty estimates for selecting active-learning points | Active learning models · Boston, Superconduct, Power | test RMSE during data collection | Improved |
| Use Combined uncertainty estimates for selecting active-learning points | Active learning models · Other UCI datasets | test RMSE during data collection | No clear effect |
| Fit decoder variance with the Combined method | VAE · Artificial four-dimensional data | generative sample quality | Improved |
Closest papers by meaning
- Estimating Regression Predictive Distributions with Sample Networks (AAAI 2023 · similarity 0.89)
- Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles (NeurIPS 2017 · similarity 0.88)
- Cooperative Variance Estimation and Bayesian Neural Networks for Disentangling Aleatoric and Epistemic Uncertainties (ICML 2026 · similarity 0.88)
- SURE: SUrvey REcipes for building reliable and robust deep networks (CVPR 2024 · similarity 0.86)
- Uncertainty Quantification with the Empirical Neural Tangent Kernel (NeurIPS 2025 · similarity 0.86)
- On the Variance of Neural Network Training with respect to Test Sets and Distributions (ICLR 2024 · similarity 0.86)
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.
