Skip to yearly menu bar Skip to main content


Dichotomy of Early and Late Phase Implicit Biases Can Provably Induce Grokking

Kaifeng Lyu · Jikai Jin · Zhiyuan Li · Simon Du · Jason Lee · Wei Hu

Halle B #129
[ ]
Fri 10 May 1:45 a.m. PDT — 3:45 a.m. PDT


Recent work by Power et al. (2022) highlighted a surprising "grokking" phenomenon in learning arithmetic tasks: a neural net first "memorizes" the training set, resulting in perfect training accuracy but near-random test accuracy, and after training for sufficiently longer, it suddenly transitions to perfect test accuracy. This paper studies the grokking phenomenon in theoretical setups and shows that it can be induced by a dichotomy of early and late phase implicit biases. Specifically, when training homogeneous neural nets with large initialization and small weight decay on both classification and regression tasks, we prove that the training process gets trapped at a solution corresponding to a kernel predictor for a long time, and then a very sharp transition to min-norm/max-margin predictors occurs, leading to a dramatic change in test accuracy.

Live content is unavailable. Log in and register to view live content