Research map / Bandits and online learning / Online convex optimization and games
SafePredict: A Meta-Algorithm for Machine Learning That Uses Refusals to Guarantee Correctness
Mustafa Kocak, David Ramirez, Elza Erkip, Dennis E. Shasha
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 |
|---|---|---|---|
| increase adaptivity alpha (weight-shifting) (ablation) | SafePredict · synthetic change-point sequences | efficiency T*/T | Improved |
| increase number of change points | SafePredict · synthetic change-point sequences | efficiency T*/T | Hurt |
| use SafePredict with alpha>0 instead of alpha=0 (ablation) | SafePredict · synthetic data | tracking of base predictor | Improved |
| raise alpha with constant base error above target (ablation) | SafePredict · synthetic data, finite sequences | validity (error rate) | Hurt |
| use SafePredict instead of confidence-based refusal | Random forest on MNIST · MNIST after change point t>5000 | validity | Improved |
| use confidence-based refusal instead of SafePredict | Random forest on MNIST · MNIST | efficiency | Improved |
| stack SafePredict on top of confidence-based refusal | Confidence-based refusal · MNIST | efficiency and validity | Improved |
| add amnesic adaptivity (forget old data when refusals exceed 0.5) (ablation) | CBR+SafePredict · MNIST after change point | efficiency | Improved |
Closest papers by meaning
- ElliCE: Efficient and Provably Robust Algorithmic Recourse via the Rashomon Sets (NeurIPS 2025 · similarity 0.83)
- Provably Reliable Conformal Prediction Sets in the Presence of Data Poisoning (ICLR 2025 · similarity 0.82)
- Algorithms with Calibrated Machine Learning Predictions (ICML 2025 · similarity 0.82)
- List Update with Prediction (AAAI 2025 · similarity 0.82)
- Overcoming Brittleness in Pareto-Optimal Learning Augmented Algorithms (NeurIPS 2024 · similarity 0.82)
- Active, anytime-valid risk controlling prediction sets (NeurIPS 2024 · similarity 0.82)
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.
