Research map / Learning theory and algorithms / Kernels and regression theory
FineMorphs: Affine-Diffeomorphic Sequences for Regression
Michele Lohr, Laurent Younes
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 |
|---|---|---|---|
| add diffeomorphic module to affine-only regression (A -> ADA) | ADA · Concrete (UCI standard splits) | test RMSE: 10.31 -> 4.93 | Improved |
| add diffeomorphic module to affine-only regression (A -> ADA) | ADA · Yacht (UCI standard splits) | test RMSE: 8.95 -> 0.72 | Improved |
| increase time steps T1 from 10 to 20 in ADA (ablation) | ADA · Airfoil | test RMSE: 1.53 -> 1.40 | Improved |
| increase time steps T1 from 20 to 40 (ablation) | ADA · Airfoil | test RMSE: 1.40 -> 1.44 | No clear effect |
| stack more AD modules with decreasing kernel widths (2 to 4) (ablation) | AD^xA (h_q decreasing) · Airfoil | test RMSE: 1.25 -> 1.18 -> 1.08 | Improved |
| use decreasing (coarse-to-fine) instead of increasing kernel widths (ablation) | AD4A · Airfoil | test RMSE: 1.32 -> 1.08 | Improved |
| stack more AD pairs (AD)^xA from 2 to 4 (ablation) | (AD)^xA · Airfoil | test RMSE: 1.41 -> 1.28 -> 1.21 | Improved |
| add a fifth AD pair (ablation) | (AD)^xA · Airfoil | test RMSE: 1.21 -> 1.25 | No clear effect |
| use AD4A (decreasing widths) instead of ADA | ADA · Concrete (UCI standard splits) | test RMSE: 4.93 -> 4.49 | Improved |
| use AD4A (decreasing widths) instead of ADA | ADA · Kin8nm (UCI standard splits) | test RMSE: 0.07 -> 0.06 | Improved |
| use AD4A (decreasing widths) instead of ADA | ADA · Power (UCI gap splits) | test RMSE: 5.53 -> 4.60 | Improved |
| use AD4A (decreasing widths) instead of ADA | ADA · Yacht (UCI standard splits) | test RMSE: 0.72 -> 0.78 | Hurt |
Closest papers by meaning
- Diffeomorphic Learning (JMLR 2020 · similarity 0.89)
- Low-Rank Tensor Transitions (LoRT) for Transferable Tensor Regression (ICML 2025 · similarity 0.84)
- Generalized Tensor Decomposition for Understanding Multi-Output Regression under Combinatorial Shifts (NeurIPS 2024 · similarity 0.84)
- Transformers can optimally learn regression mixture models (ICLR 2024 · similarity 0.84)
- TMPNN: High-Order Polynomial Regression Based on Taylor Map Factorization (AAAI 2024 · similarity 0.83)
- Affine Equivariant Networks Based on Differential Invariants (CVPR 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.
