Research map / Label-efficient and robust learning / Recognition and long-tail
Consistent Polyhedral Surrogates for Top-k Classification and Variants
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 |
|---|---|---|---|
| Construct a polyhedral surrogate embedding the top-k loss and use the argmax link | Top-k classification · Top-k classification | calibration | Helped |
| Use L_2 with the argmax link instead of a consistent top-k surrogate | Top-k classification · Top-k classification | consistency | Hurt |
| Use L_3 with the argmax link instead of a consistent top-k surrogate | Top-k classification · Top-k classification | consistency | Hurt |
| Use L_4 with the argmax link instead of a consistent top-k surrogate | Top-k classification · Top-k classification | consistency | Hurt |
| Optimize L_k rather than the inconsistent surrogates | Top-k classification with n=5 and k=3 · Dirichlet(α, α, 1, 1, 1) distributions | top-k regret | Helped |
| Optimize L_2 rather than L_k | Top-k classification with n=5 and k=3 · Dirichlet(α, α, 1, 1, 1) distributions | top-k regret | Hurt |
| Optimize L_4 rather than L_k | Top-k classification with n=5 and k=3 · Dirichlet(α, α, 1, 1, 1) distributions | top-k regret | Hurt |
| Train with L_k rather than the other surrogates | Linear model for top-k classification · Dirichlet(αp) conditional distributions at large α | top-k test loss | Helped |
| Train with L_k rather than the inconsistent surrogates | Linear model for top-k classification · Dirichlet(αp) conditional distributions at smaller α | top-k test loss | Hurt |
Closest papers by meaning
- On the consistency of top-k surrogate losses (ICML 2020 · similarity 0.91)
- Trading off Consistency and Dimensionality of Convex Surrogates for Multiclass Classification (NeurIPS 2024 · similarity 0.88)
- Cardinality-Aware Set Prediction and Top-$k$ Classification (NeurIPS 2024 · similarity 0.84)
- Consistency Conditions for Differentiable Surrogate Losses (NeurIPS 2025 · similarity 0.83)
- A Universal Growth Rate for Learning with Smooth Surrogate Losses (NeurIPS 2024 · similarity 0.83)
- Realizable $H$-Consistent and Bayes-Consistent Loss Functions for Learning to Defer (NeurIPS 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.