Research map / Label-efficient and robust learning / Recognition and long-tail
A no-regret generalization of hierarchical softmax to extreme multi-label classification
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 |
|---|---|---|---|
| Replace HSM pick-one-label training with PLT-based XT | fastText-style XMLC · Delicious-200K | P@1: 42.22 -> 47.85 | Helped |
| Replace HSM pick-one-label training with PLT-based XT | fastText-style XMLC · WikiLSHTC | P@1: 41.13 -> 58.73 | Helped |
| Replace HSM pick-one-label training with PLT-based XT | fastText-style XMLC · Wiki-500K | P@1: 32.73 -> 64.48 | Helped |
| Replace HSM pick-one-label training with PLT-based XT | fastText-style XMLC · Amazon-670K | P@1: 25.47 -> 39.90 | Helped |
| Use PLT marginal probability estimates instead of pick-one-label probabilities | HSM for multi-label classification · precision@k with conditionally dependent labels | precision@k regret | Helped |
| Use pick-one-label reduction when labels are conditionally independent | HSM for multi-label classification · precision@k with conditionally independent labels | precision@k regret | No clear effect |
| Use top-down clustering rather than a Huffman tree (ablation) | XT · WikiLSHTC, Wiki-500K, and Amazon-670K | precision@k | Mixed |
| Weight feature vectors by TF-IDF rather than simple averaging (ablation) | XT · WikiLSHTC, Wiki-500K, and Amazon-670K | precision@k | Mixed |
| Add L2 regularization (ablation) | XT · WikiLSHTC, Wiki-500K, and Amazon-670K | precision@k | Mixed |
Closest papers by meaning
- ADIOS: Architectures Deep In Output Space (ICML 2016 · similarity 0.90)
- Generalized test utilities for long-tail performance in extreme multi-label classification (NeurIPS 2023 · similarity 0.90)
- ELMO : Efficiency via Low-precision and Peak Memory Optimization in Large Output Spaces (ICML 2025 · similarity 0.88)
- Batch Selection for Multi-Label Classification Guided by Uncertainty and Dynamic Label Correlations (AAAI 2025 · similarity 0.87)
- Taming the Sigmoid Bottleneck: Provably Argmaxable Sparse Multi-Label Classification (AAAI 2024 · similarity 0.87)
- Self-Paced Unified Representation Learning for Hierarchical Multi-Label Classification (AAAI 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.