LGMay 20, 2025

New Evidence of the Two-Phase Learning Dynamics of Neural Networks

arXiv:2505.13900v13 citationsh-index: 6
Originality Incremental advance
AI Analysis

This work provides incremental insights into the fundamental challenge of how deep neural networks learn, which is important for researchers in machine learning.

The paper tackled the problem of understanding the two-phase learning dynamics of deep neural networks by introducing an interval-wise perspective, revealing the Chaos Effect and Cone Effect that show a transition from chaotic sensitivity to stable refinement during training.

Understanding how deep neural networks learn remains a fundamental challenge in modern machine learning. A growing body of evidence suggests that training dynamics undergo a distinct phase transition, yet our understanding of this transition is still incomplete. In this paper, we introduce an interval-wise perspective that compares network states across a time window, revealing two new phenomena that illuminate the two-phase nature of deep learning. i) \textbf{The Chaos Effect.} By injecting an imperceptibly small parameter perturbation at various stages, we show that the response of the network to the perturbation exhibits a transition from chaotic to stable, suggesting there is an early critical period where the network is highly sensitive to initial conditions; ii) \textbf{The Cone Effect.} Tracking the evolution of the empirical Neural Tangent Kernel (eNTK), we find that after this transition point the model's functional trajectory is confined to a narrow cone-shaped subset: while the kernel continues to change, it gets trapped into a tight angular region. Together, these effects provide a structural, dynamical view of how deep networks transition from sensitive exploration to stable refinement during training.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes