Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Multi-class Gaussian Process Classification with Noisy Inputs
Carlos Villacampa-Calvo, Bryan Zaldívar, Eduardo C. Garrido-Merchán, Daniel Hernández-Lobato
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 |
|---|---|---|---|
| model input noise (NIMGP) instead of ignoring it in the multi-class GP | MGP · 1D toy, 3 classes | test NLL: 0.884 -> 0.286 | Improved |
| model input noise on the toy problem | MGP · 1D toy, 3 classes | test error: 0.125 -> 0.129 | No clear effect |
| use neural-network-computed posterior parameters (NIMGP_NN) instead of free parameters (ablation) | NIMGP · 1D toy, 3 classes | test NLL: 0.286 -> 0.347 | Hurt |
| use full noise-aware inference (NIMGP) instead of first-order approximation (NIMGP_FO) (ablation) | NIMGP_FO · 1D toy, 3 classes | test NLL: 0.495 -> 0.286 | Improved |
| augment training data with posterior samples of inputs | MGP · Synthetic, noise 0.1 | test NLL: 0.76 -> 0.8 | No clear effect |
| sample input attributes from the posterior in each batch | MGP · Synthetic, noise 0.5 | test NLL: 1.5 -> 0.74 | Improved |
| sample input attributes from the posterior in each batch | MGP · Synthetic, noise 0.5 | test error: 0.22 -> 0.26 | Hurt |
| add NIMGP noise modelling at 7 classes and noise 0.5 | MGP · Synthetic, C=7 classes, noise 0.5 | test NLL: 3.06 -> 0.991 | Improved |
| add NIMGP noise modelling with small training set N=100, D=10, noise 0.1 | MGP · Synthetic, N=100, D=10, noise 0.1 | test NLL: 0.273 -> 0.246 | Improved |
| remove Detection and Curvature Significance attributes (ablation) | NIMGP · Astrophysics | NLL | Hurt |
| add NIMGP on astrophysics data without two attributes | MGP · Astrophysics (without Detection and Curvature Significance) | NLL: 0.381 -> 0.268 | Improved |
| add NIMGP on astrophysics data without two attributes | MGP · Astrophysics (without Detection and Curvature Significance) | test error: 0.077 -> 0.0811 | Hurt |
Closest papers by meaning
- Scalable Multi-Class Gaussian Process Classification using Expectation Propagation (ICML 2017 · similarity 0.87)
- Efficient Transformed Gaussian Processes for Non-Stationary Dependent Multi-class Classification (ICML 2023 · similarity 0.86)
- GP-Tree: A Gaussian Process Classifier for Few-Shot Incremental Learning (ICML 2021 · similarity 0.85)
- A scalable end-to-end Gaussian process adapter for irregularly sampled time series classification (NeurIPS 2016 · similarity 0.85)
- Computation-Aware Gaussian Processes: Model Selection And Linear-Time Inference (NeurIPS 2024 · similarity 0.83)
- Solving High Frequency and Multi-Scale PDEs with Gaussian Processes (ICLR 2024 · 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.
