Research map / Learning theory and algorithms / Prediction and decision losses
Lossless or Quantized Boosting with Integer Arithmetic
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 |
|---|---|---|---|
| Q-loss-based boosting (RATBOOST) instead of AdaBoost, linear weak learners | RATBOOST · UCI Hardware (Figure 1) | test error convergence over iterations | Helped |
| Adaptive weight quantization with 2-bit index (RATBOOSTA2) | RATBOOST · UCI 25 domains (Table 1 excerpt) | test error (%) | Mixed |
| Integer arithmetic version with q=11 quantization (RATBOOSTE) (ablation) | RATBOOST · UCI 25 domains (Table 1 excerpt) | test error (%) | Mixed |
| Deterministic regular 2-bit weight quantization (RATBOOSTQ2) (ablation) | RATBOOST · UCI Hardware (Figure 1) | convergence to good solutions | Hurt |
| Adaptive 2-bit quantization in place of double precision weights | RATBOOST · UCI Hardware (Figure 1) | weight coding size: ~31.9x reduction | Helped |
| Stochastic rounding instead of deterministic quantization of weights (ablation) | RATBOOST · UCI 25 domains (Table 1 excerpt) | test error (%) | Hurt |
| Bayes risk of Q-loss as decision tree splitting criterion | decision trees · theoretical (WLA convergence) | boosting convergence rate | Helped |
Closest papers by meaning
- How to Boost Any Loss Function (NeurIPS 2024 · similarity 0.88)
- Boosting with Tempered Exponential Measures (NeurIPS 2023 · similarity 0.86)
- Quantized Training of Gradient Boosting Decision Trees (NeurIPS 2022 · similarity 0.86)
- Revisiting Agnostic Boosting (NeurIPS 2025 · similarity 0.85)
- Optimal Parallelization of Boosting (NeurIPS 2024 · similarity 0.84)
- QBB: Quantization with Binary Bases for LLMs (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.