Research map / Label-efficient and robust learning / Recognition and long-tail
Multi-class Probabilistic Bounds for Majority Vote Classifiers with Partially Labeled Data
Vasilii Feofanov, Emilie Devijver, Massih-Reza Amini
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 |
|---|---|---|---|
| add MSTA self-training with automatic vote-threshold selection to the supervised random forest | Random Forest · MNIST | ACC-U: .798 -> .857 | Improved |
| add MSTA self-training to the supervised random forest | Random Forest · Isolet | ACC-U: .854 -> .875 | Improved |
| add MSTA self-training to the supervised random forest | Random Forest · Pendigits | ACC-U: .863 -> .884 | Improved |
| add MSTA self-training to the supervised random forest | Random Forest · Vowel | ACC-U: .586 -> .586 | No clear effect |
| add MSTA self-training to the supervised random forest | Random Forest · SensIT | ACC-U: .723 -> .722 | No clear effect |
| choose the pseudo-labeling threshold automatically (MSTA) instead of fixing theta=0.7 (FSTA) | FSTA · MNIST | ACC-U: .705 -> .857 | Improved |
| choose the pseudo-labeling threshold automatically (MSTA) instead of curriculum percentile (CSTA) | CSTA · Isolet | ACC-U: .843 -> .875 | Improved |
| use MSTA instead of QN-S3VM with many classes | QN-S3VM · Letter | ACC-U: .064 -> .717 | Improved |
| use QN-S3VM instead of MSTA on DNA | MSTA · DNA | ACC-U: .702 -> .815 | Hurt |
| use Semi-LDA instead of MSTA on Protein | MSTA · Protein | ACC-U: .781 -> .842 | Hurt |
| increase the share of labeled examples until sufficient | MSTA vs RF · MNIST subsample (3500) | accuracy | No clear effect |
| use label spreading and QN-S3VM on large datasets | LS, QN-S3VM · Fashion, MNIST, SensIT | feasibility | Hurt |
Closest papers by meaning
- Transductive Bounds for the Multi-Class Majority Vote Classifier (AAAI 2019 · similarity 0.95)
- Erratum: Risk Bounds for the Majority Vote: From a PAC-Bayesian Analysis to a Learning Algorithm (JMLR 2023 · similarity 0.89)
- Second Order PAC-Bayesian Bounds for the Weighted Majority Vote (NeurIPS 2020 · similarity 0.89)
- On Margins and Generalisation for Voting Classifiers (NeurIPS 2022 · similarity 0.88)
- How many classifiers do we need? (NeurIPS 2024 · similarity 0.86)
- Don’t Label Twice: Quantity Beats Quality when Comparing Binary Classifiers on a Budget (ICML 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.
