2020 (modified: 12 May 2023)ICML 2020Readers: Everyone
Abstract:We study the implicit regularization of mini-batch stochastic gradient descent, when applied to the fundamental problem of least squares regression. We leverage a continuous-time stochastic differe...