Provable Benefit of Cutout and CutMix for Feature Learning

Documenti analoghi

• From Linear to Nonlinear: Provable Weak-to-Strong Generalization through Feature Learning
  di: Oh, Junsoo, et al.
  Pubblicazione: (2025)
• DASH: Warm-Starting Neural Network Training in Stationary Settings without Loss of Plasticity
  di: Shin, Baekrok, et al.
  Pubblicazione: (2024)
• Through the River: Understanding the Benefit of Schedule-Free Methods for Language Model Training
  di: Song, Minhak, et al.
  Pubblicazione: (2025)
• Minor First, Major Last: A Depth-Induced Implicit Bias of Sharpness-Aware Minimization
  di: Moon, Chaewon, et al.
  Pubblicazione: (2026)
• The Cost of Robustness: Tighter Bounds on Parameter Complexity for Robust Memorization in ReLU Nets
  di: Kim, Yujun, et al.
  Pubblicazione: (2025)