Research map / Probabilistic and scientific modeling / Molecules and proteins
Hierarchical and Stochastic Crystallization Learning: Geometrically Leveraged Nonparametric Regression with Delaunay Triangulation
Jiaqi Gu, Guosheng Yin
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 |
|---|---|---|---|
| use crystallization learning instead of local linear regression | k-NN/LL baselines · Scenario 1, d=5, n=2000 | log(MSE) ratio: 1.06 | Improved |
| use crystallization learning instead of 5-NN in high dimension | 5-NN · Scenario 1, d=50, n=2000 | log(MSE) ratio: 0.17 | No clear effect |
| increase sample size n from 200 to 2000 (ablation) | crystallization learning · Scenario 1, d=5 | log(MSE): -1.11 -> -2.21 | Improved |
| use crystallization learning under jump points of feature density | crystallization learning · Scenario 2, d=5, n=2000 | log(MSE) ratio vs kernel regression: 1.53 | Improved |
| add data-driven leave-one-out selection of L (ablation) | crystallization learning · Scenario 1, d=5 | log(MSE) | Improved |
| use stochastic instead of deterministic crystallization (ablation) | crystallization learning · synthetic, n=200, d=5 | log(MSE) | Improved |
| use hierarchical crystallization learning instead of GAM, Gaussian density model | GAM · Gaussian density additive model, d=5 | log(MSE ratio): 0.65 -> 1.07 | Improved |
| use hierarchical crystallization learning instead of GAM, Sine model | GAM · Sine additive model, d=5 | log(MSE ratio): 0.06 -> 0.09 | No clear effect |
| increase number of representative points n* from 500 to 2000 (ablation) | hierarchical crystallization learning · YearPredictionMSD | log(MPSE ratio vs GAM): 0.08 -> 0.20 | Improved |
| use stochastic crystallization learning instead of deterministic (ablation) | crystallization learning · UCI real data | log(MPSE) | Improved |
Closest papers by meaning
- Crystallization Learning with the Delaunay Triangulation (ICML 2021 · similarity 0.96)
- Nonparametric Regression on Random Geometric Graphs Sampled from Submanifolds (JMLR 2025 · similarity 0.84)
- BAST: Bayesian Additive Regression Spanning Trees for Complex Constrained Domain (NeurIPS 2021 · similarity 0.84)
- Winner-takes-all learners are geometry-aware conditional density estimators (ICML 2024 · similarity 0.84)
- Diffusion Model's Generalization Can Be Characterized by Inductive Biases toward a Data-Dependent Ridge Manifold (NeurIPS 2026 · similarity 0.83)
- Bridging Multicalibration and Out-of-distribution Generalization Beyond Covariate Shift (NeurIPS 2024 · similarity 0.83)
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.
