Research map / Probabilistic and scientific modeling / Dynamical systems
Identifiable Smooth Conjugacy Learning via Adversarial Orthogonality
In Huh, Changwook Jeong, Muhammad Alam
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 adversarial orthogonality regularizer to context modulation (ablation) | CM conjugacy · Hopf normal form | dH: 0.142 -> 0.081 | Improved |
| add adversarial orthogonality to context modulation | CM conjugacy · Hopf observed space | dH: 0.064 -> 0.064 | No clear effect |
| allow source correction (CM, Aug., DT) instead of fixed source | baseline conjugacy · Hopf observed space | dH: 0.547 -> 0.064 | Improved |
| add orthogonality regularizer | CM conjugacy · Duffing oscillator | Z2 symmetry / test MSE | Improved |
| add orthogonality regularizer (ablation) | CM conjugacy · Hopf extrapolation | dH (normal form): 0.177 -> 0.103 | Improved |
| add orthogonality regularizer (ablation) | CM conjugacy · FitzHugh-Nagumo omega=2 canonical | MSE: 0.703 -> 0.218 | Improved |
| use direct tuning or L2-SP of source weights | conjugacy learning · FitzHugh-Nagumo omega=1 canonical | MSE: 2.199 -> 199.0 | Hurt |
| use less expressive NICE adversary instead of Real NVP (ablation) | CM-AO · Hopf normal form | dH: 0.081 -> 0.093 | Hurt |
| use CNF adversary instead of Real NVP (ablation) | CM-AO · Hopf normal form | dH: 0.081 -> 0.075 | No clear effect |
| increase lambda_orth to 1e-2 (ablation) | CM-AO · Hopf normal form | MMD: 0.109 -> 0.304 | Hurt |
| use hinge-loss variant at lambda_orth 1e-2 (ablation) | CM-AO · Hopf normal form | MMD: 0.304 -> 0.106 | Improved |
| moderate lambda_orth between 1e-4 and 3e-3 (ablation) | CM-AO · Hopf normal form | dH: 0.142 -> 0.071 | Improved |
Closest papers by meaning
- Identifying Equivalent Training Dynamics (NeurIPS 2024 · similarity 0.88)
- Beyond DSA: Conjugacy-based Comparison of Dynamical Systems (NeurIPS 2026 · similarity 0.88)
- Extracting Governing Equations from Latent Dynamics via Multi-View Contrastive Learning (NeurIPS 2026 · similarity 0.86)
- End-to-End Identifiable and Consistent Recurrent Switching Dynamical Systems (NeurIPS 2026 · similarity 0.86)
- Out-of-Domain Generalization in Dynamical Systems Reconstruction (ICML 2024 · similarity 0.86)
- Identifying latent state transitions in non-linear dynamical systems (ICLR 2025 · 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.
