Skip to yearly menu bar Skip to main content


The Complexity of Finding Stationary Points with Stochastic Gradient Descent

Yoel Drori · Ohad Shamir

Keywords: [ Information Theory and Estimation ] [ Non-convex Optimization ] [ Optimization - Non-convex ]

Abstract: We study the iteration complexity of stochastic gradient descent (SGD) for minimizing the gradient norm of smooth, possibly nonconvex functions. We provide several results, implying that the classical $\mathcal{O}(\epsilon^{-4})$ upper bound (for making the average gradient norm less than $\epsilon$) cannot be improved upon, unless a combination of additional assumptions is made. Notably, this holds even if we limit ourselves to convex quadratic functions. We also show that for nonconvex functions, the feasibility of minimizing gradients with SGD is surprisingly sensitive to the choice of optimality criteria.

Chat is not available.