New algorithm HT-PAder tackles online convex optimization with heavy-tailed noise

Researchers have developed HT-PAder, a novel parameter-free algorithm designed to tackle online convex optimization challenges in non-stationary environments with heavy-tailed noise. This new approach combines restarted AdaGrad experts with a pathwise meta-algorithm, AdaGrad-Hedge, which bypasses the need for moment conditions on meta-losses. HT-PAder achieves a near-optimal universal dynamic regret, outperforming previous methods even in scenarios with finite variance noise and without requiring prior knowledge of problem parameters. AI

IMPACT Introduces a new algorithmic approach for optimization problems common in machine learning.

RANK_REASON Academic paper detailing a new algorithm for online convex optimization. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• AdaGrad
• AdaGrad-Hedge
• arXiv
• HT-PAder

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →