Research map / Learning theory and algorithms / Kernels and regression theory
A Statistical Mechanics Framework for Task-Agnostic Sample Design in Machine Learning
Bhavya Kailkhura, Jayaraman Thiagarajan, Qunwei Li, Jize Zhang, Yi Zhou, Timo Bremer
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 |
|---|---|---|---|
| train on blue noise sample design instead of random sampling | 2-hidden-layer MLP (200,100) · Synthetic exponential function, 3D | test RMSE | Improved |
| train on Poisson disk sampling design instead of random sampling | 2-hidden-layer MLP (200,100) · Synthetic exponential function, 3D | test RMSE | Improved |
| train on blue noise design instead of random sampling | 2-hidden-layer MLP · Synthetic disk function, 3D | test RMSE | Improved |
| train on Poisson disk design instead of random sampling | 2-hidden-layer MLP · Synthetic disk function, 3D | test RMSE | Improved |
| train on blue noise design instead of random sampling | 2-hidden-layer MLP · ICF simulator peak fusion power | test RMSE | Improved |
| train on blue noise design instead of random sampling | 2-hidden-layer MLP · ICF simulator radiation energy | test RMSE | Improved |
| train on Poisson disk design instead of random sampling | 2-hidden-layer MLP · ICF simulator (peak fusion power, radiation energy) | test RMSE | Improved |
| use spectral designs in the low-sample regime versus larger sample sizes (200 to 1000) | blue noise / PDS designs · Synthetic and ICF regression | gain over random design | Improved |
| blue noise with large zero region when function bandwidth is below the zero-region limit | blue noise sampler · Best-case band-limited function (theory) | generalization gap | Improved |
| blue noise when function bandwidth exceeds rho_z* | blue noise sampler · Best and worst-case functions (theory) | convergence rate of generalization gap | No clear effect |
Closest papers by meaning
- A Spectral Approach for the Design of Experiments: Design, Analysis and Algorithms (JMLR 2018 · similarity 0.85)
- Sharpened Generalization Bounds based on Conditional Mutual Information and an Application to Noisy, Iterative Algorithms (NeurIPS 2020 · similarity 0.85)
- Structural-Entropy-Based Sample Selection for Efficient and Effective Learning (ICLR 2025 · similarity 0.84)
- Generalization Bounds for Gradient Methods via Discrete and Continuous Prior (NeurIPS 2022 · similarity 0.84)
- Interpretability and Generalization Bounds for Learning Spatial Physics (ICML 2026 · similarity 0.83)
- On Generalization Error Bounds of Noisy Gradient Methods for Non-Convex Learning (ICLR 2020 · 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.
