Research map / Learning theory and algorithms / Prediction and decision losses
DNNR: Differential Nearest Neighbors Regression
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 DNNR instead of KNN | Regression · eight UCI benchmark datasets | MSE | Helped |
| use local gradient-based Taylor predictions rather than simple target averaging | KNN regression · PMLB | median R²: over 0.1 | Helped |
| add feature scaling to DNNR (ablation) | DNNR · Airfoil and Concrete | MSE: 4.82 ± 0.74 -> 2.83 ± 0.60; 49.97 ± 13.59 -> 36.52 ± 18.03 | Helped |
| add feature scaling to DNNR (ablation) | DNNR · Friedman-1 | MSE: 1.03 ± 0.15 -> 0.01 ± 0.00 | Helped |
| use second-order Taylor approximation (ablation) | DNNR · Airfoil and Concrete | MSE: 2.83 ± 0.60 -> 2.30 ± 0.48; 36.52 ± 18.03 -> 28.35 ± 13.47 | Helped |
| use second-order Taylor approximation (ablation) | DNNR · California, CO2 Emission, and NOx Emission | MSE: 0.19 ± 0.02 -> 0.22 ± 0.02; 1.12 ± 0.47 -> 1.24 ± 0.50; 13.34 ± 0.96 -> 15.61 ± 1.94 | Hurt |
| use second-order Taylor approximation (ablation) | DNNR · Protein and Yacht | MSE: 12.31 ± 0.35 -> 11.64 ± 0.44; 1.05 ± 0.63 -> 0.48 ± 0.43 | Helped |
| add MLKR scaling to KNN averaging (ablation) | KNN · Airfoil, Concrete, and Friedman-1 | MSE: 4.25 ± 0.96 -> 3.32 ± 0.84; 60.76 ± 14.19 -> 36.85 ± 9.89; 3.93 ± 0.43 -> 0.40 ± 0.07 | Helped |
| replace DNNR scaling with MLKR (ablation) | DNNR · Concrete | MSE: 36.52 ± 18.03 -> 1.5e13 ± 4e13 | Hurt |
| replace DNNR scaling with MLKR (ablation) | DNNR · Airfoil | MSE: 2.83 ± 0.60 -> 3.20 ± 1.08 | Helped |
| use randomly selected neighbors instead of nearest neighbors (ablation) | DNNR · Airfoil, Concrete, and Friedman-1 | MSE: 2.83 ± 0.60 -> 24.38 ± 3.44; 36.52 ± 18.03 -> 115.14 ± 20.52; 0.01 ± 0.00 -> 6.07 ± 0.56 | Hurt |
| add Lasso regularization to gradient estimation (ablation) | DNNR · Airfoil, Concrete, and Friedman-1 | MSE: 2.83 ± 0.60 -> 5.25 ± 1.05; 36.52 ± 18.03 -> 40.24 ± 8.43; 0.01 ± 0.00 -> 7.33 ± 0.52 | Hurt |
Closest papers by meaning
- k*-Nearest Neighbors: From Global to Local (NeurIPS 2016 · similarity 0.89)
- On Convergence of Nearest Neighbor Classifiers over Feature Transformations (NeurIPS 2020 · similarity 0.89)
- Rates of Convergence for Large-scale Nearest Neighbor Classification (NeurIPS 2019 · similarity 0.87)
- Revisiting Nearest Neighbor for Tabular Data: A Deep Tabular Baseline Two Decades Later (ICLR 2025 · similarity 0.85)
- Feature Averaging: An Implicit Bias of Gradient Descent Leading to Non-Robustness in Neural Networks (ICLR 2025 · similarity 0.84)
- Supervised Kernel Thinning (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.