phase transition
Phase transition in machine learning refers to a sudden change in the behavior or performance of a model or algorithm as a parameter (e.g., model size, data quantity) crosses a critical threshold, affecting generalization and learning dynamics.
- Bounds on the computational complexity of neurons due to dendritic morphology
- Do Language Models Use Their Depth Efficiently?
- Exponential Dynamic Energy Network for High Capacity Sequence Memory
- From Information to Generative Exponent: Learning Rate Induces Phase Transitions in SGD
- Generalization vs Specialization under Concept Shift
- On Traceability in $\ell_p$ Stochastic Convex Optimization
- On the Edge of Memorization in Diffusion Models
- The $\varphi$ Curve: The Shape of Generalization through the Lens of Norm-based Capacity Control
- The Price of Sparsity: Sufficient Conditions for Sparse Recovery using Sparse and Sparsified Measurements