Study finds finite-width effects shift the edge of chaos in deep neural network initialisation
A new arXiv paper shows that finite network width shifts the edge of chaos, the critical line of weight and bias variances where deep neural networks train best, away from its infinite-width prediction.
Read the full story at arXiv stat.ML ↗
Timeline · 1 report
- 2026-11-05 05:00 · arXiv stat.ML
Finite-width shift of the edge of chaos in deep neural networks