Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Practical Equivariances via Relational Conditional Neural Processes
Daolang Huang, Manuel Haussmann, Ulpu Remes, S. T. John, Grégoire Clarté, Kevin Sebastian Luck, Samuel Kaski, Luigi Acerbi
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 |
|---|---|---|---|
| Replace CNP encoder with relational encoding (RCNP, stationary) | CNP · Weakly-periodic GP d_x=1, OOID | score: 2.88 -> 0.24 | Improved |
| Replace GNP with relational RGNP (stationary) | GNP · Mixture d_x=5, OOID | log-likelihood: -0.72 -> 0.04 | Improved |
| Replace CNP with RCNP (stationary) | CNP · Mixture d_x=5, OOID | log-likelihood: -1.71 -> -0.31 | Improved |
| Replace CNP with RGNP (stationary) | CNP · Sawtooth d_x=5, OOID | log-likelihood: -0.22 -> 1.13 | Improved |
| Use isotropic comparison (FullRCNP) instead of stationary (RCNP) | RCNP · EQ GP isotropic d_x=5, OOID | score: 0.45 -> 0.35 | Improved |
| Use isotropic comparison (FullRGNP) instead of stationary (RGNP) | RGNP · EQ GP isotropic d_x=5, OOID | score: 0.11 -> 0.25 | Hurt |
| Replace CNP with RCNP (stationary) | CNP · EQ GP isotropic d_x=5, OOID | score: 1.22 -> 0.45 | Improved |
| Use RGNP instead of ConvGNP on higher-dimensional tasks | ConvGNP · Weakly-periodic d_x=3, 5 | applicability | Improved |
| Replace ConvGNP with RGNP (stationary) | ConvGNP · Weakly-periodic GP d_x=1, OOID | score: 0.01 -> 0.03 | No clear effect |
| Replace ConvGNP with RGNP (stationary) | ConvGNP · Lotka-Volterra, forecasting, simulated | normalized log-likelihood: -4.30 -> -4.27 | Improved |
| Replace ConvGNP with RGNP (stationary) | ConvGNP · Lotka-Volterra, interpolation, simulated | normalized log-likelihood: -3.46 -> -3.51 | Hurt |
| Replace CNP with RGNP (stationary) | CNP · Reaction-Diffusion, completion | normalized log-likelihood: 0.13 -> 1.38 | Improved |
Closest papers by meaning
- Approximately Equivariant Neural Processes (NeurIPS 2024 · similarity 0.86)
- Translation Equivariant Transformer Neural Processes (ICML 2024 · similarity 0.86)
- Spectral Convolutional Conditional Neural Process (NeurIPS 2025 · similarity 0.86)
- Distance-informed Neural Processes (NeurIPS 2025 · similarity 0.85)
- Neural Conditional Probability for Uncertainty Quantification (NeurIPS 2024 · similarity 0.84)
- Rényi Neural Processes (ICML 2025 · similarity 0.84)
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.
