Research map / Probabilistic and scientific modeling / Molecules and proteins
From Predictions to Decisions: Using Lookahead Regularization
Nir Rosenfeld, Anna Hilgard, Sai Srivatsa Ravindranath, David C. Parkes
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 lookahead regularization (small step size) | quadratic regression · Quadratic curves, eta=0.75 | mean improvement magnitude: 1.109 -> 1.108 | No clear effect |
| add lookahead regularization (moderate step size) | quadratic regression · Quadratic curves, eta=1.25 | mean improvement rate: 0.143 -> 0.714 | Improved |
| add lookahead regularization (moderate step size) | quadratic regression · Quadratic curves, eta=1.25 | RMSE: 0.342 -> 0.424 | Hurt |
| add lookahead regularization (moderate step size) | quadratic regression · Quadratic curves, eta=1.25 | mean improvement magnitude: -0.261 -> 1.065 | Improved |
| add lookahead regularization (very large step size) | quadratic regression · Quadratic curves, eta=3.5 | mean improvement magnitude: -35.13 -> 0.604 | Mixed |
| increase lookahead lambda under full mutability | linear model · Wine quality, all features mutable (eta=0.5) | improvement vs RMSE | Mixed |
| increase ridge alpha for baseline | linear ridge model · Wine quality, full mutability | improvement | Hurt |
| increase lookahead lambda under partial mutability | linear model · Wine quality, partial mutability (eta=2) | improvement with RMSE | Improved |
| use lookahead instead of baseline under partial mutability | linear model · Wine quality, partial mutability | improvement rate: ~40% -> 100% | Improved |
| add lookahead on diabetes with linear f and bootstrapped g | linear model · Diabetes, BMI and s1 mutable (eta=5) | improvement rate and magnitude with accuracy | Improved |
| use quadratic f with quantile regression uncertainty | quadratic model · Diabetes non-linear setting | individualized decisions toward low uncertainty | Improved |
| lookahead regularization when x' moves far from the sample distribution (eta=3.5) | lookahead model · Quadratic curves, eta=3.5 | mean improvement rate: 0 -> 0.571 | Mixed |
Closest papers by meaning
- On the Actionability of Outcome Prediction (AAAI 2024 · similarity 0.86)
- Look-Ahead Reasoning on Learning Platforms (NeurIPS 2025 · similarity 0.84)
- Performative Prediction with Selective Labels (NeurIPS 2026 · similarity 0.84)
- Decision Theoretic Foundations for Conformal Prediction: Optimal Uncertainty Quantification for Risk-Averse Agents (ICML 2025 · similarity 0.84)
- Regretful Decisions under Label Noise (ICLR 2025 · similarity 0.84)
- Performative Prediction with Bandit Feedback: Learning through Reparameterization (ICML 2024 · similarity 0.84)
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.
