Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
April 15, 2026Journal of Global OptimizationOpen Access

Quasi-Newton method with subspace gradients

View Full Paper
Ask AI
Bookmark
Share

Authors

TMTaisei MiyaishiThe University of TokyoRNRyota NozawaThe University of TokyoPPPierre‐Louis PoirionRIKEN Center for Advanced Intelligence Project

Discussion

Loading...

Member takes

Implication

This algorithm improves iteration efficiency in large-scale optimization, suggesting enhanced computation time benefits.

Key Points

  • The aim is to develop a subspace quasi-Newton method that avoids the computation of full gradients and Hessians while maintaining efficiency.
  • Developed a quasi-Newton method using a deterministic subspace.
  • Utilized subspace gradients based on random matrix theory.
  • Evaluated performance through numerical experiments comparing computation times.
  • Achieved similar worst-case iteration complexity as existing subspace methods.
  • Demonstrated superior computation time in numerical experiments.

Cite This Study

Miyaishi et al. (2026) studied this question.

synapsesocial.com/papers/69df2a99e4eeef8a2a6af992https://doi.org/10.1007/s10898-026-01597-7
View Full Paper
Ask AI
Bookmark
Share

Also Consider

Synapse has enriched one closely related paper. Consider it for comparative context:

  1. 1Analysis of a Symmetric Rank-One Trust Region Method1996 · 64 citations