Skip to yearly menu bar Skip to main content


Quadratic models for understanding catapult dynamics of neural networks

Libin Zhu · Chaoyue Liu · Adityanarayanan Radhakrishnan · Misha Belkin

Halle B #199
[ ]
Thu 9 May 1:45 a.m. PDT — 3:45 a.m. PDT


While neural networks can be approximated by linear models as their width increases, certain properties of wide neural networks cannot be captured by linear models. In this work we show that recently proposed Neural Quadratic Models can exhibit the "catapult phase" Lewkowycz et al. (2020) that arises when training such models with large learning rates. We then empirically show that the behaviour of quadratic models parallels that of neural networks in generalization, especially in the catapult phase regime. Our analysis further demonstrates that quadratic models are an effective tool for analysis of neural networks.

Chat is not available.