Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Beyond the Mean-Field: Structured Deep Gaussian Processes Improve the Predictive Uncertainties
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 |
|---|---|---|---|
| Allow inducing outputs to covary across processes and layers rather than using mean-field independence (ablation) | Fully-coupled DGP · UCI benchmark datasets | ELBO | Helped |
| Use the stripes-and-arrow covariance structure instead of a fully coupled covariance (ablation) | Structured DGP · Protein UCI runtime | runtime: three times faster | Helped |
| Use stripes-and-arrow covariance rather than mean-field independence (ablation) | DGP · Extrapolation across UCI datasets | marginal test log-likelihood comparison frequency: 5/8 datasets | Helped |
| Use stripes-and-arrow covariance rather than mean-field independence (ablation) | DGP · Interpolation on UCI datasets | marginal test log-likelihood | No clear effect |
| Analytically marginalise inducing outputs instead of Monte Carlo sampling them (ablation) | Fully-coupled DGP · Concrete UCI | ELBO convergence | Helped |
| Use mean-field inference instead of stripes-and-arrow covariance (ablation) | DGP · Kin8nm calibration under extrapolation | uncertainty calibration | Hurt |
| Use stripes-and-arrow covariance instead of mean-field independence (ablation) | DGP · Kin8nm predictive uncertainty | uncertainty calibration | Helped |
Closest papers by meaning
- Amortized Structured Stochastic Variational Inference for Gaussian Process Latent Variable Models (NeurIPS 2026 · similarity 0.92)
- Structured and Efficient Variational Deep Learning with Matrix Gaussian Posteriors (ICML 2016 · similarity 0.92)
- Variational Inference for Gaussian Process Models with Linear Complexity (NeurIPS 2017 · similarity 0.92)
- Deep Gaussian Processes with Importance-Weighted Variational Inference (ICML 2019 · similarity 0.91)
- Neighbour-Driven Gaussian Process Variational Autoencoders for Scalable Structured Latent Modelling (ICML 2025 · similarity 0.90)
- Bayesian Basis Function Approximation for Scalable Gaussian Process Priors in Deep Generative Models (ICML 2025 · similarity 0.90)
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.