Research map / Learning theory and algorithms / Combinatorial and search optimization
Stabilizing Equation Learning via Zero-Point Constraints
Sannyuya Liu, Ao Chen, Lin Liu, Ruxia Liang, Xiaoxuan Shen, Jianwen Sun
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 zero-point constraint and zero-point consistent operators (ablation) | EQL-Z · Nguyen (12 functions) | exact recovery rate (avg): 86.3% -> 98.3% | Improved |
| add small-to-large structure search instead of a single large fixed network (ablation) | EQL-Z · Nguyen (12 functions) | exact recovery rate (avg): 17.9% -> 98.3% | Improved |
| apply EQL-Z (zero-point, small-to-large search, BFGS) to vanilla EQL | EQL · Nguyen (12 functions) | exact recovery rate (avg): 0.0% -> 98.3% | Improved |
| apply EQL-Z to vanilla EQL | EQL · Oscillation 1 | OOD NMSE: 2.3894 -> 0.0087 | Improved |
| apply EQL-Z to vanilla EQL | EQL · Stress-strain | OOD NMSE: 1.0349 -> 0.1597 | Improved |
| apply EQL-Z to vanilla EQL | EQL · E. coli growth | ID NMSE: 0.9709 -> 0.8405 | No clear effect |
| enforce Lambda(0)=0 on operators (ablation) | EQL · (x1+x2)^2 toy task | exp operator gradient: peak > 0.6 -> zero | Improved |
| apply EQL-Z to vanilla EQL | EQL · SRBench black-box | average model complexity: reduced by more than 80% | Improved |
| use EQL-Z instead of GP-based methods | EQL-Z · SRBench black-box | R2 | Hurt |
| add BFGS fine-tuning of coefficients | EQL-Z · Nguyen | R2 | Improved |
| language-model priors (LLM-SR) versus EQL-Z | LLM-SR · real-world tasks | ID/OOD NMSE | Improved |
Closest papers by meaning
- Deep Generative Symbolic Regression (ICLR 2023 · similarity 0.88)
- Beyond Accuracy and Complexity: The Effective Information Criterion for Structurally Stable Symbolic Regression (ICML 2026 · similarity 0.88)
- SMILE: Bridging Continuous Optimization and Discrete Symbolic Recovery (NeurIPS 2026 · similarity 0.87)
- Symbolic regression via MDLformer-guided search: from minimizing prediction error to minimizing description length (ICLR 2025 · similarity 0.86)
- ODEFormer: Symbolic Regression of Dynamical Systems with Transformers (ICLR 2024 · similarity 0.85)
- Symmetry-Informed Governing Equation Discovery (NeurIPS 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.
