Research map / Learning theory and algorithms / Prediction and decision losses
Quantifying Learning Guarantees for Convex but Inconsistent Surrogates
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 |
|---|---|---|---|
| Use the generalized calibration-function lower bound for quadratic surrogates without requiring the score space to contain the loss-matrix column space | Quadratic surrogate with an arbitrary score subspace · General finite-output prediction | calibration-function lower bound | Helped |
| Increase tree consistency depth, allowing scores to distinguish more labels | Quadratic surrogate with tree-structured loss · Tree-structured multi-class classification | best achievable task-loss accuracy | Helped |
| Increase tree consistency depth | Quadratic surrogate with tree-structured loss · Tree-structured multi-class classification | calibration-function quadratic coefficient | Hurt |
| Restrict mAP scores to the sorting score matrix instead of using the full mAP score matrix | Quadratic surrogate for mAP ranking · mAP ranking | prediction computational complexity | Helped |
| Restrict mAP scores to the sorting score matrix instead of using the full mAP score matrix | Quadratic surrogate for mAP ranking · mAP ranking | level of inconsistency | Hurt |
Closest papers by meaning
- Bayes Consistency vs. H-Consistency: The Interplay between Surrogate Loss Functions and the Scoring Function Class (NeurIPS 2020 · similarity 0.92)
- Trading off Consistency and Dimensionality of Convex Surrogates for Multiclass Classification (NeurIPS 2024 · similarity 0.92)
- Realizable $H$-Consistent and Bayes-Consistent Loss Functions for Learning to Defer (NeurIPS 2024 · similarity 0.91)
- A Universal Growth Rate for Learning with Smooth Surrogate Losses (NeurIPS 2024 · similarity 0.90)
- Consistency Conditions for Differentiable Surrogate Losses (NeurIPS 2025 · similarity 0.88)
- $H$-Consistency Guarantees for Regression (ICML 2024 · similarity 0.87)
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.