Can We Remove the Square-Root in Adaptive Gradient Methods? A Second-Order Perspective

Documenti analoghi

• Structured Inverse-Free Natural Gradient: Memory-Efficient & Numerically-Stable KFAC
  di: Lin, Wu, et al.
  Pubblicazione: (2023)
• Spectral-factorized Positive-definite Curvature Learning for NN Training
  di: Lin, Wu, et al.
  Pubblicazione: (2025)
• Fully Adaptive Zeroth-Order Method for Minimizing Functions with Compressible Gradients
  di: Grapiglia, Geovani Nunes, et al.
  Pubblicazione: (2025)
• Square Root-Factorized Covariance Steering
  di: Kumagai, Naoya, et al.
  Pubblicazione: (2026)
• Convergence of First-Order Algorithms with Momentum from the Perspective of an Inexact Gradient Descent Method
  di: Khanh, Pham Duy, et al.
  Pubblicazione: (2025)