Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Random Feature Expansions 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 |
|---|---|---|---|
| Use RBF rather than arc-cosine covariance | Random-feature DGP · Airline | accuracy: 78.55% vs 72.76% | Helped |
| Use RBF rather than arc-cosine covariance | Random-feature DGP · Airline | MNLL: 0.4583 vs 0.5335 | Helped |
| Reintroduce original inputs in the hidden layer (ablation) | DGP-RBF · MNIST | accuracy: 98.04% -> 98.2% | Helped |
| Increase depth from two layers to 10 or 20 layers, feeding original inputs to each hidden layer (ablation) | DGP-RBF · Airline | error rate | Helped |
| Use variational spectral frequencies with fixed randomness rather than prior-fixed or resampled frequencies (ablation) | Random-feature DGP · EEG and UCI datasets | accuracy | Helped |
| Resample variational spectral-frequency randomness at each iteration (ablation) | Random-feature DGP · EEG | accuracy | Hurt |
| Use arc-cosine covariance with variational frequencies and fixed randomness on higher-dimensional data (ablation) | DGP-arc · Spam and MNIST | generalization | Hurt |
| Increase the number of random features (ablation) | Random-feature DGP · EEG | generalization | Mixed |
| Use more than two hidden layers without feeding original inputs forward (ablation) | DGP-RBF · MNIST | accuracy | Hurt |
Closest papers by meaning
- Doubly Stochastic Variational Inference for Deep Gaussian Processes (NeurIPS 2017 · similarity 0.91)
- Deep Gaussian Processes for Regression using Approximate Expectation Propagation (ICML 2016 · similarity 0.90)
- Deep Gaussian Processes with Importance-Weighted Variational Inference (ICML 2019 · similarity 0.90)
- Beyond the Mean-Field: Structured Deep Gaussian Processes Improve the Predictive Uncertainties (NeurIPS 2020 · similarity 0.89)
- Inference in Deep Gaussian Processes using Stochastic Gradient Hamiltonian Monte Carlo (NeurIPS 2018 · similarity 0.89)
- Approximate Inference Turns Deep Networks into Gaussian Processes (NeurIPS 2019 · 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.