Research map / Probabilistic and scientific modeling / Molecules and proteins
Dimensionless machine learning: Imposing exact units equivariance
Soledad Villar, Weichi Yao, David W. Hogg, Ben Blum-Smith, Bianca Dumitrascu
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 dimensionless scalar inputs (units equivariance) for the Hamiltonian MLP | Scalar-based Hamiltonian MLP · Springy double pendulum, Experiment 2 (kg inputs rescaled) | state relative error: .3669 -> .0089 | Improved |
| use dimensionless scalar inputs (units equivariance) for the Hamiltonian MLP | Scalar-based Hamiltonian MLP · Springy double pendulum, Experiment 3 (shifted parameter distribution) | state relative error: .1885 -> .0435 | Improved |
| use dimensionless scalar inputs (units equivariance) for the Hamiltonian MLP | Scalar-based Hamiltonian MLP · Springy double pendulum, Experiment 1 (in-distribution) | state relative error: .0055 -> .0061 | No clear effect |
| use dimensionless features in linear regression | Rietkerk vegetation linear regression · Rietkerk vegetation regression (held-out test set) | test MSE: 26.3 -> 12.6 | Improved |
| use dimensionless features in linear regression | Rietkerk vegetation linear regression · Rietkerk vegetation regression (held-out test set) | Pearson correlation: 0.94 -> 0.97 | Improved |
| augment the 286 dimensionless monomials with 500 random dimensional monomials (ablation) | Hamiltonian symbolic regression (L2 and LASSO) · Springy pendulum Hamiltonian symbolic regression | held-out mean-squared error: an order of magnitude worse mean-squared error | Hurt |
| restrict features to dimensionless monomials | Hamiltonian symbolic regression · Springy pendulum Hamiltonian symbolic regression | number of candidate monomials: 187 500 -> 286 | Improved |
| use only dimensionless features when a needed dimensional constant (Planck's h) is missing | Black-body radiation regression · Planck black-body law | fit quality | Hurt |
Closest papers by meaning
- Approximately Equivariant Neural Processes (NeurIPS 2024 · similarity 0.87)
- Neural Fourier Transform: A General Approach to Equivariant Representation Learning (ICLR 2024 · similarity 0.86)
- Equivariance via Minimal Frame Averaging for More Symmetries and Efficiency (ICML 2024 · similarity 0.86)
- Probing Equivariance and Symmetry Breaking in Convolutional Networks (NeurIPS 2025 · similarity 0.86)
- Equivariance by Contrast: Identifiable Equivariant Embeddings from Unlabeled Finite Group Actions (NeurIPS 2025 · similarity 0.85)
- Emergent Equivariance in Deep Ensembles (ICML 2024 · similarity 0.85)
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.
