Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Fully Bayesian Autoencoders with Latent Sparse Gaussian Processes
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 |
|---|---|---|---|
| Replace variational inference with fully Bayesian inference using SGHMC | Bayesian autoencoder with a GP latent prior · Rotated MNIST digit 3 | MSE: 0.0257 $\pm$ 0.0004 -> 0.0228 $\pm$ 0.0004 | Helped |
| Replace variational inference with fully Bayesian inference using SGHMC | Bayesian autoencoder with a GP latent prior · EEG | SMSE: 0.52 $\pm$ 0.05 -> 0.22 $\pm$ 0.01 | Helped |
| Replace variational inference with fully Bayesian inference using SGHMC | Bayesian autoencoder with a GP latent prior · Jura | NLL: 1.02 $\pm$ 0.01 -> 0.91 $\pm$ 0.04 | Helped |
| Replace a single GP prior with a deep GP prior (ablation) | SGP-BAE · EEG | SMSE: 0.22 $\pm$ 0.01 -> 0.21 $\pm$ 0.01 | Helped |
| Replace a single GP prior with a deep GP prior (ablation) | SGP-BAE · EEG | NLL: 1.96 $\pm$ 0.08 -> 2.25 $\pm$ 0.13 | Hurt |
| Replace a single GP prior with a deep GP prior (ablation) | SGP-BAE · Jura | MAE: 0.45 $\pm$ 0.03 -> 0.44 $\pm$ 0.02 | Helped |
| Replace a single GP prior with a deep GP prior (ablation) | SGP-BAE · Jura | NLL: 0.91 $\pm$ 0.04 -> 0.85 $\pm$ 0.04 | Helped |
| Use a fully Bayesian treatment of inducing inputs and GP hyperparameters instead of point estimates | Sparse GP autoencoder · Moving ball | RMSE | Helped |
| Use a correlated GP prior instead of an isotropic Gaussian latent prior | Bayesian autoencoder · Moving ball | trajectory reconstruction | Helped |
| Use a correlated GP prior instead of an independent-output model | Missing-data imputation · EEG and Jura | prediction performance | Helped |
Closest papers by meaning
- Model Selection for Bayesian Autoencoders (NeurIPS 2021 · similarity 0.89)
- Langevin Autoencoders for Learning Deep Latent Variable Models (NeurIPS 2022 · similarity 0.89)
- High-Dimensional Bayesian Optimisation with Gaussian Process Prior Variational Autoencoders (ICLR 2025 · similarity 0.89)
- Bayesian Basis Function Approximation for Scalable Gaussian Process Priors in Deep Generative Models (ICML 2025 · similarity 0.89)
- Neighbour-Driven Gaussian Process Variational Autoencoders for Scalable Structured Latent Modelling (ICML 2025 · similarity 0.88)
- PG-LBO: Enhancing High-Dimensional Bayesian Optimization with Pseudo-Label and Gaussian Process Guidance (AAAI 2024 · 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.