Remove Symmetries to Control Model Expressivity and Improve Optimization

Documenti analoghi

• Parameter Symmetry Potentially Unifies Deep Learning Theory
  di: Ziyin, Liu, et al.
  Pubblicazione: (2025)
• Three Mechanisms of Feature Learning in a Linear Network
  di: Xu, Yizhou, et al.
  Pubblicazione: (2024)
• Thermodynamic Irreversibility of Training Algorithms
  di: Ziyin, Liu, et al.
  Pubblicazione: (2026)
• Does SGD Seek Flatness or Sharpness? An Exactly Solvable Model
  di: Xu, Yizhou, et al.
  Pubblicazione: (2026)
• Neural Thermodynamics: Entropic Forces in Deep and Universal Representation Learning
  di: Ziyin, Liu, et al.
  Pubblicazione: (2025)