Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Deep Gaussian Processes with Importance-Weighted Variational Inference
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 importance-weighted inference instead of variational inference (ablation) | Deep latent-variable Gaussian processes · UCI datasets (41 datasets) | median difference from GP baseline: 0.12 -> 0.32 | Helped |
| Add latent variables to deep GP mappings | Deep Gaussian processes · UCI datasets (41 datasets) | median difference from GP baseline: 0.06 -> 0.12 | Helped |
| Use a third GP layer in the latent-variable model with IWVI | LV-GP-GP · UCI datasets (41 datasets) | median difference from GP baseline: 0.26 -> 0.32 | Helped |
| Use IWVI instead of VI (ablation) | LV-GP-GP-GP · servo | average test log-likelihood: -0.25 (0.01) -> 0.03 (0.03) | Helped |
| Use IWVI instead of VI (ablation) | LV-GP-GP · autompg | average test log-likelihood: -0.38 (0.04) -> -0.24 (0.01) | Helped |
| Use IWVI instead of VI (ablation) | LV-GP-GP · keggundirected | average test log-likelihood: -64 (6) -> 2.37 (0.29) | Helped |
| Use IWVI instead of VI (ablation) | LV-GP-GP-GP · winered | average test log-likelihood: -1.06 (0.03) -> -1.31 (0.40) | Hurt |
| Add latent variables to a deep GP | GP-GP on forest · forest | average test log-likelihood: -1.37 (0.00) -> -0.91 (0.00) | Helped |
| Add latent variables to a deep GP | GP-GP-GP on bike · bike | average test log-likelihood: 3.68 (0.01) -> 3.95 (0.01) | Helped |
| Use a noise-free deep GP instead of a latent-variable deep GP | kin40k · kin40k | average test log-likelihood: 1.17 (0.00) -> 1.27 (0.00) | Hurt |
Closest papers by meaning
- On Signal-to-Noise Ratio Issues in Variational Inference for Deep Gaussian Processes (ICML 2021 · similarity 0.91)
- Doubly Stochastic Variational Inference for Deep Gaussian Processes (NeurIPS 2017 · similarity 0.91)
- Sparse Inducing Points in Deep Gaussian Processes: Enhancing Modeling with Denoising Diffusion Variational Inference (ICML 2024 · similarity 0.88)
- New Bounds for Sparse Variational Gaussian Processes (ICML 2025 · similarity 0.88)
- Forward $\chi^2$ Divergence Based Variational Importance Sampling (ICLR 2024 · similarity 0.87)
- Amortized Structured Stochastic Variational Inference for Gaussian Process Latent Variable Models (NeurIPS 2026 · 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.