Research map / Learning theory and algorithms / Combinatorial and search optimization
Reinforcement-Enhanced Autoregressive Feature Transformation: Gradient-steered Search in Continuous Space for Postfix Expressions
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| replace RL-based data collection with random collection (MOAT-d) (ablation) | MOAT · SpectF, SpamBase, Openml_616, Openml_618 | F1-score / 1-RAE | Helped |
| add data augmentation by shuffling segments (remove it: MOAT-a) (ablation) | MOAT · SpectF, SpamBase, Openml_616, Openml_618 | F1-score / 1-RAE | Helped |
| beam size 5 instead of 1 (ablation) | MOAT · 23 datasets | valid rate | Helped |
| RL-collected data and postfix beam search instead of random records and greedy search (DIFER) | MOAT · SpectF | F1-score: 0.766 -> 0.912 | Helped |
| MOAT vs second-best GRFG on Higgs Boson | MOAT · Higgs Boson | F1-score: 0.707 -> 0.712 | Helped |
| MOAT vs GRFG on Openml_616 | MOAT · Openml_616 | 1-RAE: 0.372 -> 0.526 | Helped |
| MOAT vs GRFG on Openml_637 | MOAT · Openml_637 | 1-RAE: 0.307 -> 0.465 | Helped |
| change downstream model from RF to Ridge | MOAT · SpectF | F1-score: 0.912 -> 0.780 | Hurt |
| MOAT vs best baseline with XGB downstream | MOAT · SpectF | F1-score: 0.842 -> 0.897 | Helped |
| MOAT vs best baselines with DT downstream | MOAT · SpectF | F1-score: 0.864 -> 0.929 | Helped |
| larger search step size eta (ablation) | MOAT · SpectF | precision/recall/F1 | Mixed |
| vary trade-off alpha 0.05 to 0.50 and eta 1 to 10 (ablation) | MOAT · SpectF | F1-score: standard deviation lower than 0.01 | No clear effect |
Closest papers by meaning
- Sculpting Features from Noise: Reward-Guided Hierarchical Diffusion for Task-Optimal Feature Transformation (NeurIPS 2025 · similarity 0.89)
- SiT: Symmetry-invariant Transformers for Generalisation in Reinforcement Learning (ICML 2024 · similarity 0.84)
- STEP: Learning STructured Embeddings for Progressive Time Series (NeurIPS 2026 · similarity 0.83)
- VEDIT: Latent Prediction Architecture For Procedural Video Representation Learning (ICLR 2025 · similarity 0.83)
- QueST: Self-Supervised Skill Abstractions for Learning Continuous Control (NeurIPS 2024 · similarity 0.83)
- Foundation Policies with Hilbert Representations (ICML 2024 · 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.