Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
The Kernel Interaction Trick: Fast Bayesian Discovery of Pairwise Interactions in High Dimensions
Raj Agrawal, Brian Trippe, Jonathan Huggins, Tamara Broderick
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 |
|---|---|---|---|
| marginalize out theta with kernel interaction trick instead of Woodbury/naive | SKIM sampler · synthetic, N=50 | runtime and memory of p(D|tau) | Improved |
| marginalize theta (KIS) instead of jointly sampling theta and tau (FULL) | SKIM with NUTS · synthetic | NUTS runtime | Improved |
| KIS instead of FULL joint sampling | SKIM · synthetic | Gelman-Rubin R-hat mixing: greater than 1.05, with some reaching as high as 1.5 -> less than 1.05 | Improved |
| SKIM-KIS versus pairs LASSO and hierarchical LASSO | SKIM-KIS · 36 synthetic datasets | main effect FDR | Improved |
| SKIM-KIS versus LASSO methods | SKIM-KIS · 36 synthetic datasets | MSE to true effects | Improved |
| SKIM-KIS runtime versus LASSO methods | SKIM-KIS · synthetic | runtime | No clear effect |
| SKIM versus PLASSO and HLASSO on correlated covariates | SKIM · Residential Building | main effects MSE: 5.0 -> 0.1 | Improved |
| SKIM versus HLASSO | SKIM · Residential Building | pairwise effects MSE: 7.8 -> 7.0 | Improved |
| SKIM versus PLASSO and HLASSO with 100 noise covariates | SKIM · Auto MPG | fake pairwise effects selected: 4 : 99 -> 1 : 0 | Improved |
| SKIM versus HLASSO with 200 noise covariates | SKIM · Auto MPG | fake main effects selected: 6 : 46 -> 3 : 0 | Improved |
Closest papers by meaning
- The SKIM-FA Kernel: High-Dimensional Variable Selection and Nonlinear Interaction Discovery in Linear Time (JMLR 2023 · similarity 0.91)
- Interaction Measures, Partition Lattices and Kernel Tests for High-Order Interactions (NeurIPS 2023 · similarity 0.89)
- Permutation-Free High-Order Interaction Tests (ICML 2025 · similarity 0.85)
- Automated Kernel Discovery Towards Understanding High-dimensional Bayesian Optimization (NeurIPS 2026 · similarity 0.85)
- Proxy-Based Approximation of Shapley and Banzhaf Interactions (NeurIPS 2026 · similarity 0.85)
- Continuous Bayesian Model Selection for Multivariate Causal Discovery (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.
