New optimization method IBPL$^+$-TP shows superior performance in machine learning tasks

Researchers have developed a new optimization method called IBPL$^+$-TP, designed to tackle complex multiblock nonconvex nonsmooth optimization problems. This method introduces a novel two-phase adaptive momentum strategy to enhance convergence speed and allows for independent extrapolation parameters. The paper demonstrates that IBPL$^+$-TP ensures monotonic convergence of the objective function and converges to a critical point, with proven convergence rates. Its effectiveness is showcased through applications in machine learning, specifically in sparse nonnegative matrix factorization and CP decomposition, where it outperformed existing state-of-the-art methods in numerical experiments. AI

RANK_REASON The cluster contains a single academic paper detailing a new optimization method and its application to machine learning problems. [lever_c_demoted from research: ic=1 ai=0.7]

Read on arXiv cs.LG →

• alphaXiv
• arXiv
• CatalyzeX
• DagsHub
• Gotit.pub
• Hugging Face
• IBPL$^+$-TP
• ScienceCast

• paper
• other

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