Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Characterizing Deep Gaussian Processes via Nonlinear Recurrence Systems
Anh Tong, Jaesik Choi
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| Constrain squared-exponential kernel hyperparameters to avoid pathological regions | Zero-mean DGP · MNIST | classification accuracy: ≈ 10% -> 91.21% | Improved |
| Constrain squared-exponential kernel hyperparameters to avoid pathological regions | Zero-mean DGP · Boston housing and diabetes | RMSE | Improved |
| Use a spectral mixture kernel instead of a kernel with a pathological parameter region | Deep Gaussian Process · Deep-layer output distances | RMSD | Improved |
| Increase the layer width so that m > ℓ²/σ² | Zero-mean DGP with squared-exponential kernel · Deep-layer output distances | RMSD | Improved |
| Set cosine-kernel hyperparameters so that π²σ² / p² > 1 | DGP with cosine kernel · Deep-layer output distances | RMSD | Improved |
| Use nonzero-mean rather than zero-mean layers | Deep Gaussian Process · Deep-layer output distances | expected squared output distance | Improved |
| Connect each layer to the input | DGP with squared-exponential kernel · Deep-layer output distances | expected squared output distance | Improved |
Closest papers by meaning
- Gaussian Process Behaviour in Wide Deep Neural Networks (ICLR 2018 · similarity 0.89)
- Learning Scalable Deep Kernels with Recurrent Structure (JMLR 2017 · similarity 0.88)
- Deep kernel processes (ICML 2021 · similarity 0.87)
- Avoiding Kernel Fixed Points: Computing with ELU and GELU Infinite Networks (AAAI 2021 · similarity 0.86)
- Wide Feedforward or Recurrent Neural Networks of Any Architecture are Gaussian Processes (NeurIPS 2019 · similarity 0.86)
- Recurrent Memory for Online Interdomain Gaussian Processes (NeurIPS 2025 · similarity 0.86)
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.
