Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Revisiting Active Sets for Gaussian Process Decoders
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 Bayesian SAS-GP instead of Bayesian GP-LVM | Bayesian GP decoder · MNIST, 2-dim latent space | classification accuracy: 0.18 ± 0.033 -> 0.63 ± 0.022 | Helped |
| Use Bayesian SAS-GP instead of Bayesian GP-LVM | Bayesian GP decoder · Fashion-MNIST, 2-dim latent space | classification accuracy: 0.24 ± 0.043 -> 0.63 ± 0.020 | Helped |
| Use Bayesian SAS-GP instead of VAE | 2-dim latent representation · MNIST | classification accuracy: 0.54 ± 0.026 -> 0.63 ± 0.022 | Helped |
| Use Bayesian SAS-GP instead of VAE | 2-dim latent representation · Fashion-MNIST | classification accuracy: 0.58±0.008 -> 0.63 ± 0.020 | Helped |
| Increase active set size from A=100 to A=400 (ablation) | SAS-GP · MNIST | RMSE: 2.55 ± 0.98 -> 2.41 ± 0.93 | Helped |
| Increase active set size from A=100 to A=400 (ablation) | Bayesian SAS-GP · MNIST | RMSE: 2.16 ± 0.02 -> 1.99 ± 0.02 | Helped |
| Increase active set size from A=100 to A=400 (ablation) | SAS-GP · Fashion-MNIST | RMSE: 2.37 ± 0.95 -> 2.25 ± 0.90 | Helped |
| Increase active set size from A=100 to A=400 (ablation) | Bayesian SAS-GP · Fashion-MNIST | RMSE: 1.99 ± 0.17 -> 1.85 ± 0.13 | Helped |
| Increase active set size from A=100 to A=400 (ablation) | SAS-GP · CIFAR-10 | RMSE: 2.66 ± 1.08 -> 2.55 ± 1.03 | Helped |
| Increase active set size from A=100 to A=400 (ablation) | Bayesian SAS-GP · CIFAR-10 | RMSE: 2.74 ± 1.07 -> 2.57 ± 1.02 | Helped |
| Increase active set size from A=100 to A=400 (ablation) | SAS-GP · CIFAR-10 | NLPD: 3.20 ± 1.55 -> 3.32 ± 1.89 | Hurt |
| Omit the predictive-score term and keep only log p(x_A | z_A) (ablation) | SAS-GP · MNIST and Fashion-MNIST | training performance | Hurt |
Closest papers by meaning
- Incremental Variational Sparse Gaussian Process Regression (NeurIPS 2016 · similarity 0.84)
- Memory-Based Dual Gaussian Processes for Sequential Learning (ICML 2023 · similarity 0.84)
- Orthogonally Decoupled Variational Gaussian Processes (NeurIPS 2018 · similarity 0.84)
- Scalable Training of Inference Networks for Gaussian-Process Models (ICML 2019 · similarity 0.84)
- Neighbour-Driven Gaussian Process Variational Autoencoders for Scalable Structured Latent Modelling (ICML 2025 · similarity 0.83)
- Sparse Gaussian Processes: Structured Approximations and Power-EP Revisited (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.