Researchers have developed new methods, scalar and diagonal versions of VR-Adam-SHANG, to improve gradient accuracy in smooth stochastic convex optimization. These methods combine STORM-type recursive estimation, Adam-style preconditioning, and SHANG's accelerated interpolation and correction, requiring minimal gradient evaluations per iteration. The theoretical analysis shows polynomial-step convergence and improved expected objective-gap bounds. Experimental results indicate that these new methods achieve superior performance compared to existing approaches at matched sample-gradient budgets. AI
RANK_REASON The cluster contains a research paper detailing new optimization methods. [lever_c_demoted from research: ic=1 ai=1.0]
Read on Hugging Face Daily Papers →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →