Generalization in Deep Learning
It addresses a foundational problem in machine learning theory for researchers, but is incremental as it builds on existing theoretical work.
The paper tackles the open question of why deep learning generalizes well despite its complexity and instability, providing theoretical insights and discussing approaches for non-vacuous generalization guarantees.
This paper provides theoretical insights into why and how deep learning can generalize well, despite its large capacity, complexity, possible algorithmic instability, nonrobustness, and sharp minima, responding to an open question in the literature. We also discuss approaches to provide non-vacuous generalization guarantees for deep learning. Based on theoretical observations, we propose new open problems and discuss the limitations of our results.