Research map / Learning theory and algorithms / Prediction and decision losses
Beyond Accuracy and Complexity: The Effective Information Criterion for Structurally Stable Symbolic Regression
Zihan Yu, Guanren Wang, Jingtao Ding, Huandong Wang, Yong Li
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 EIC-guided rejection sampling to pretraining formulas | E2ESR · SRBench white-box | R2 / samples: 0.5190 -> 0.5399; 35M -> 9.79M samples | Improved |
| add EIC-guided rejection sampling to pretraining formulas | SNIP · SRBench white-box | R2 / samples: 0.5300 -> 0.5415; 40M -> 25.15M samples | Improved |
| add EIC-guided rejection sampling to pretraining formulas | SR4MDL · SRBench white-box | MAE / RMSE: 6.972 -> 6.812 MAE; 8.731 -> 8.701 RMSE | Improved |
| filter training formulas with EIC < 2 | random formula generator · Feynman | JS divergence #Variables: 0.1196 -> 0.0768 | Improved |
| filter training formulas with EIC < 2 | random formula generator · Strogatz | KL divergence #Variables: 18.784 -> 12.920 | Improved |
| filter training formulas with EIC < 2 | random formula generator · Wiki Named Equations | KL divergence #Coefficients: 14.888 -> 3.5307 | Improved |
| train E2ESR on EIC-filtered formulas for same number of steps | E2ESR · Feynman test | R2: 0.55 -> 0.68 | Improved |
| add EIC to fitness of GP and MCTS | GP and MCTS · SRBench white-box | Pareto tier: second -> first Pareto tier | Improved |
| add EIC to fitness of GP and MCTS | GP and MCTS · SRBench black-box (122) | Pareto tier: GP Tier 2 -> 1; MCTS Tier 4 -> 3 | Improved |
| select lower-EIC formula among pairs of equal accuracy and complexity | expert judgement · Expert preference study | preference rate: 69.8% (95% CI: [67%, 73%]) | Improved |
Closest papers by meaning
- Contemporary Symbolic Regression Methods and their Relative Performance (NeurIPS 2021 · similarity 0.88)
- SMILE: Bridging Continuous Optimization and Discrete Symbolic Recovery (NeurIPS 2026 · similarity 0.88)
- AI Feynman 2.0: Pareto-optimal symbolic regression exploiting graph modularity (NeurIPS 2020 · similarity 0.88)
- Pareto-Optimal Fronts for Benchmarking Symbolic Regression Algorithms (ICML 2025 · similarity 0.87)
- FunctionEvolve: Structure-Guided Symbolic Regression with LLMs (NeurIPS 2026 · similarity 0.86)
- Ab Initio Nonparametric Variable Selection for Scalable Symbolic Regression with Large $p$ (ICML 2025 · similarity 0.86)
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.
