Research map / Label-efficient and robust learning / Recognition and long-tail
Revisiting F-measure Optimization in Multi-Label Classification: A Sampling-based Approach
Zixun Wang
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 |
|---|---|---|---|
| replace direct multinomial estimation with sampling-based estimation | Bayes rule for F-measure · bibtex | F1 (%): 37.85 -> 44.98 | Improved |
| replace direct multinomial estimation with sampling-based estimation | Bayes rule for F-measure · medical | F1 (%): 38.54 -> 53.81 | Improved |
| replace direct multinomial estimation with sampling-based estimation | Bayes rule for F-measure · COCO | F1 (%): 77.27 -> 78.24 | Improved |
| replace matrix multiplication with FFT convolutions | Lambda computation · COCO | time (s): 1.35 -> 1.03 | Improved |
| replace matrix multiplication with FFT convolutions | Lambda computation · yeast | time (s): 0.20 -> 0.10 | Improved |
| use sampling-based estimation instead of direct multinomial estimation | V estimation runtime · COCO | time (s): 85.10 -> 101.74 | Hurt |
| use Gibbs sampling instead of autoregressive sampling (ablation) | SMN · six MLC datasets | F1 (%): medical 53.81 -> 56.68; bibtex 44.98 -> 44.67 | No clear effect |
| use random label orders instead of default order (ablation) | SMN autoregressive sampling · six MLC datasets | F1 (%): medical 53.81 -> 53.89 (mean) | No clear effect |
| use Bayes-rule predictions per metric from one SMN model vs separate models | SMN · yeast | F1 (%): 62.20 -> 66.02 | Improved |
| use asymmetric loss to handle label imbalance instead of DMN | ASL vs DMN · bibtex | F1 (%): 37.85 -> 38.33 | Improved |
| use Convex Calibrated Surrogate with independent binary estimators instead of DMN | CCS · medical | F1 (%): 38.54 -> 23.94 | Hurt |
Closest papers by meaning
- Convex Calibrated Surrogates for the Multi-Label F-Measure (ICML 2020 · similarity 0.92)
- Consistent algorithms for multi-label classification with macro-at-$k$ metrics (ICLR 2024 · similarity 0.86)
- Taming the Sigmoid Bottleneck: Provably Argmaxable Sparse Multi-Label Classification (AAAI 2024 · similarity 0.85)
- Generalization Analysis for Multi-Label Learning (ICML 2024 · similarity 0.85)
- Estimation of the Label-Noise Transition Matrix with Performance Guarantees via Selective Classification (NeurIPS 2026 · similarity 0.85)
- Predicting Label Distribution from Ternary Labels (NeurIPS 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.
