Research map / Probabilistic and scientific modeling / Variational and generative sampling
On Signal-to-Noise Ratio Issues in Variational Inference for Deep 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 REG with DREG | Two-layer latent-variable DGP · forest | test log-likelihood: 0.59 -> 0.63 | Helped |
| Replace REG with DREG | Two-layer latent-variable DGP · solar | test log-likelihood: 2.33 -> 2.57 | Helped |
| Replace REG with DREG | Two-layer latent-variable DGP · pol | test log-likelihood: 2.99 -> 2.99 | No clear effect |
| Replace REG with DREG | Two-layer latent-variable DGP · power | test log-likelihood: 0.21 -> 0.21 | No clear effect |
| Replace REG with DREG | Two-layer latent-variable DGP · winewhite | test log-likelihood: -1.11 -> -1.11 | No clear effect |
| Replace REG with DREG | Two-layer latent-variable DGP · winered | test log-likelihood: 0.57 -> 0.61 | No clear effect |
| Increase the number of importance samples K | REG latent-variable gradient estimator in DGP-IWVI · Real-world datasets and DGP depths | gradient SNR: O(√(M/K)) | Hurt |
| Use DREG instead of REG | Latent-variable gradient estimator in DGP-IWVI · Real-world datasets and DGP depths | gradient SNR: O(1/√K) -> O(√K) | Helped |
Closest papers by meaning
- Deep Gaussian Processes with Importance-Weighted Variational Inference (ICML 2019 · similarity 0.91)
- Doubly Stochastic Variational Inference for Deep Gaussian Processes (NeurIPS 2017 · similarity 0.89)
- Beyond the Mean-Field: Structured Deep Gaussian Processes Improve the Predictive Uncertainties (NeurIPS 2020 · similarity 0.89)
- Sticking the Landing: Simple, Lower-Variance Gradient Estimators for Variational Inference (NeurIPS 2017 · similarity 0.88)
- Sparse Inducing Points in Deep Gaussian Processes: Enhancing Modeling with Denoising Diffusion Variational Inference (ICML 2024 · similarity 0.88)
- Forward $\chi^2$ Divergence Based Variational Importance Sampling (ICLR 2024 · similarity 0.88)
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.