Data Uniformity Improves Training Efficiency and More, with a Convergence Framework Beyond the NTK Regime

Documenti analoghi

• LoRA Training in the NTK Regime has No Spurious Local Minima
  di: Jang, Uijeong, et al.
  Pubblicazione: (2024)
• Federated Dynamical Low-Rank Training with Global Loss Convergence Guarantees
  di: Schotthöfer, Steffen, et al.
  Pubblicazione: (2024)
• Implicit Bias of Per-sample Adam on Separable Data: Departure from the Full-batch Regime
  di: Baek, Beomhan, et al.
  Pubblicazione: (2025)
• Global Convergence and Rich Feature Learning in $L$-Layer Infinite-Width Neural Networks under $μ$P Parametrization
  di: Chen, Zixiang, et al.
  Pubblicazione: (2025)
• New Hybrid Fine-Tuning Paradigm for LLMs: Algorithm Design and Convergence Analysis Framework
  di: Ma, Shaocong, et al.
  Pubblicazione: (2026)