Research map / Learning theory and algorithms / Kernels and regression theory
Topologically penalized regression on manifolds
Olympio Hacquard, Krishnakumar Balasubramanian, Gilles Blanchard, Clément Levrard, Wolfgang Polonik
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 Lasso with the persistence penalty Omega_2 | Laplacian eigenbasis regression · Torus (n=1000, sigma 1) | RMSE: 0.290 -> 0.209 | Improved |
| Replace Lasso with persistence-weighted Lasso Omega_1 | Laplacian eigenbasis regression · Torus (n=300, sigma 1) | RMSE: 0.381 -> 0.281 | Improved |
| Replace total variation penalty with Omega_2 | manifold regression · Swiss roll (sigma 1) | RMSE: 0.920 -> 0.546 | Improved |
| Replace KRR with Omega_2 at low noise | manifold regression · Swiss roll (sigma 0.2) | RMSE: 0.186 -> 0.398 | Hurt |
| Replace KRR with Omega_2 at high noise | manifold regression · Swiss roll (sigma 1.3) | RMSE: 0.746 -> 0.595 | Improved |
| Replace KRR with Omega_2 | spinning cat 1D prediction · Spinning cat 1D (sigma 0, 68/4) | RMSE: 0.22 -> 0.35 | Hurt |
| Replace Lasso with Omega_2 | spinning cat 1D prediction · Spinning cat 1D (sigma 0, 68/4) | RMSE: 0.79 -> 0.35 | Improved |
| Replace KRR with Omega_2 | spinning cat 2D prediction · Spinning cat 2D (train/test 800/200) | RMSE: 0.290 -> 0.273 | Improved |
| Reduce training set from 108 to 58 samples with Omega_2 | electrical consumption prediction · Electrical consumption | RMSE: 1.251 -> 1.165 | Improved |
| Add Omega_1 variable selection before Omega_2 (ablation) | Omega_2 · Torus and Swiss roll | RMSE | Improved |
| Increase p from 100 to 1000 in Omega_1 (n=1000) | Omega_1 · Torus computation time | seconds: 14.2 -> 132.3 | Hurt |
| Use Omega_2 instead of Lasso in compute time | Omega_2 · Torus computation time (n=1000, p=100) | seconds: 0.19 -> 129.9 | Hurt |
Closest papers by meaning
- Sample complexity and effective dimension for regression on manifolds (NeurIPS 2020 · similarity 0.90)
- Multivariate Regression with Gross Errors on Manifold-Valued Data (TPAMI 2018 · similarity 0.88)
- Learning Globally Smooth Functions on Manifolds (ICML 2023 · similarity 0.87)
- The Noisy Laplacian: a Threshold Phenomenon for Non-Linear Dimension Reduction (ICML 2025 · similarity 0.86)
- Diffeomorphic interpolation for efficient persistence-based topological optimization (NeurIPS 2024 · similarity 0.84)
- Fréchet Regression on the Bures-Wasserstein Manifold (NeurIPS 2026 · 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.
