Research map / Learning theory and algorithms / Prediction and decision losses
Generalized Boosting
Arun Suggala, Bingbin Liu, Pradeep K. Ravikumar
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 dense connections to earlier layers (DenseCompBoost vs StdCompBoost) (ablation) | compositional boosting · Simulation 3 | test accuracy: 82.49 -> 95.70 | Improved |
| Use dense connections (DenseCompBoost vs StdCompBoost) (ablation) | compositional boosting · Simulation 1 | test accuracy: 91.53 -> 93.55 | Improved |
| Increase complexity of weak transformer with depth (CmplxCompBoost vs StdCompBoost) (ablation) | compositional boosting · Simulation 1 | test accuracy: 91.53 -> 91.97 | Improved |
| Increase complexity of weak transformer with depth (CmplxCompBoost vs StdCompBoost) (ablation) | compositional boosting · CIFAR10 | test accuracy: 81.93 -> 82.43 | No clear effect |
| Compositional boosting instead of additive feature boosting | StdCompBoost · SVHN | test accuracy: 83.36 -> 90.81 | Improved |
| Compositional boosting instead of additive feature boosting | StdCompBoost · Simulation 3 | test accuracy: 73.13 -> 82.49 | Improved |
| Compositional boosting instead of XGBoost trees | DenseCompBoost · CIFAR10 | test accuracy: 58.34 -> 82.31 | Improved |
| DenseCompBoost instead of XGBoost on Simulation 3 | DenseCompBoost · Simulation 3 | test accuracy: 50.10 -> 95.70 | Improved |
| Use DenseCompBoost instead of XGBoost on Letter | DenseCompBoost · Letter | test accuracy: 96.16 -> 96.96 | No clear effect |
| Use DenseCompBoost instead of XGBoost on CovType | DenseCompBoost · CovType | test accuracy: 97.46 -> 96.28 | Hurt |
| Use CmplxCompBoost instead of End-to-End training | CmplxCompBoost · CIFAR10 | test accuracy: 86.88 -> 82.43 | Hurt |
| Use DenseCompBoost instead of End-to-End training | DenseCompBoost · FashionMNIST | test accuracy: 93.49 -> 93.17 | No clear effect |
Closest papers by meaning
- Sample-Efficient Agnostic Boosting (NeurIPS 2024 · similarity 0.89)
- Revisiting Agnostic Boosting (NeurIPS 2025 · similarity 0.89)
- The Many Faces of Optimal Weak-to-Strong Learning (NeurIPS 2024 · similarity 0.88)
- Weak-to-Strong Generalization Even in Random Feature Networks, Provably (ICML 2025 · similarity 0.85)
- Sample-Optimal Agnostic Boosting with Unlabeled Data (ICML 2025 · similarity 0.85)
- Improving Generalization via Meta-Learning on Hard Samples (CVPR 2024 · similarity 0.85)
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.
