The Benefits of Reusing Batches for Gradient Descent in Two-Layer Networks: Breaking the Curse of Information and Leap Exponents

Documenti analoghi

• The Computational Advantage of Depth: Learning High-Dimensional Hierarchical Functions with Gradient Descent
  di: Dandi, Yatin, et al.
  Pubblicazione: (2025)
• Online Learning and Information Exponents: On The Importance of Batch size, and Time/Complexity Tradeoffs
  di: Arnaboldi, Luca, et al.
  Pubblicazione: (2024)
• Fundamental limits of learning in sequence multi-index models and deep attention networks: High-dimensional asymptotics and sharp thresholds
  di: Troiani, Emanuele, et al.
  Pubblicazione: (2025)
• Repetita Iuvant: Data Repetition Allows SGD to Learn High-Dimensional Multi-Index Functions
  di: Arnaboldi, Luca, et al.
  Pubblicazione: (2024)
• Fundamental computational limits of weak learnability in high-dimensional multi-index models
  di: Troiani, Emanuele, et al.
  Pubblicazione: (2024)