Research map / Label-efficient and robust learning / Recognition and long-tail
Information-theoretic Classification Accuracy: A Criterion that Guides Data-driven Combination of Ambiguous Outcome Labels in Multi-class Classification
Chihao Zhang, Yiling Elaine Chen, Shihua Zhang, Jingyi Jessica Li
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 |
|---|---|---|---|
| use ITCA instead of ACC as class combination criterion | class combination (LDA) · Simulated K0=8 | # successes: 6/127 -> 120/127 | Improved |
| use ITCA instead of PE criterion | class combination (LDA) · Simulated K0=8 | # successes: 101/127 -> 120/127 | Improved |
| use ITCA instead of MI criterion with RF | class combination (RF) · Simulated K0=8 | average Hamming distance: 2.33 -> 0.08 | Improved |
| use greedy search instead of exhaustive search (ablation) | ITCA (LDA) · Simulated K0=8 | # class combinations examined: 127.00 -> 22.52 | No clear effect |
| add pruning of search space to greedy search (ablation) | ITCA greedy search · Simulated K0=8 | # class combinations examined: 22.52 -> 11.91 | Improved |
| add pruning of search space to BFS (ablation) | ITCA BFS · Simulated K0=8 | # class combinations examined: 53.61 -> 27.20 | Improved |
| use ITCA instead of ACC on Iris with LDA | class combination · Iris (K0=4) | finds true combination | Improved |
| use ITCA instead of clustering-based combination | class combination (SVM) · 4 simulated 2D datasets | finds true combination | Improved |
| add ITCA-guided interval combination for NN (ablation) | NN classifier · TCGA GBM survival | Kendall's tau: 0.6547 -> 0.6855 | Improved |
| add ITCA-guided interval combination for LR (ablation) | LR classifier · TCGA GBM survival | Kendall's tau: 0.6024 -> 0.6236 | Improved |
| use NN with ITCA-guided intervals instead of Cox regression | survival prediction · TCGA GBM survival | Kendall's tau: 0.6303 -> 0.6855 | Improved |
| reduce number of classes K from 12 downward under ITCA guidance (ablation) | NN on GBM · TCGA GBM survival | ITCA and Kendall's tau | Mixed |
Closest papers by meaning
- Multi-label classification: do Hamming loss and subset accuracy really conflict with each other? (NeurIPS 2020 · similarity 0.84)
- On Truthing Issues in Supervised Classification (JMLR 2024 · similarity 0.83)
- Data-IQ: Characterizing subgroups with heterogeneous outcomes in tabular data (NeurIPS 2022 · similarity 0.83)
- Learning Compact Semantic Information and Reliable Pseudo-Labels for Incomplete Multi-View Multi-Label Classification (TPAMI 2026 · similarity 0.83)
- Rethinking Consistent Multi-Label Classification under Inexact Supervision (ICLR 2026 · similarity 0.83)
- Rethinking and Reweighting the Univariate Losses for Multi-Label Ranking: Consistency and Generalization (NeurIPS 2021 · 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.
