Online boosting algorithm achieves multiaccuracy with residuals orthogonal to weak class
Aaroth · x · 2026-08-14
Aaroth describes how the online boosting algorithm makes probabilistic forecasts that are multiaccurate with respect to the weak class, meaning residuals are orthogonal. If error is high, ex-post reweighting by residual error yields a smooth distribution where no weak learner has non-trivial edge.
Related event: New Online Boosting Algorithm Ditches Ensembles, Runs 50x Faster(3 posts)→
More from Research
- First Agent Memory Leaderboard launches; MemoraX tops commercial text-memory track — rohanpaul_ai · 2026-08-14
- Anthropic Frontier Red Team Report: Multi-Agent Systems Face Cooperation and Trust Issues — dhadfieldmenell · 2026-08-14
- Microsoft folds Normuon row-norm into row selection for modest gains — JohnCLangford · 2026-08-14
- Scale learning rate by 1/sqrt(f) when sampling rows, Microsoft finds — JohnCLangford · 2026-08-14
- Microsoft Open-Sources Dion Optimizer: Integrates Gram-NS, Row Selection, and More, 6x Faster Training — JohnCLangford · 2026-08-14
- RL in Big Worlds workshop happening tomorrow at RL Conference — tw_killian · 2026-08-14