Stochastic Nested Variance Reduction for Nonconvex Optimization
This work addresses optimization efficiency for machine learning practitioners by offering incremental improvements in gradient complexity for nonconvex problems.
The paper tackles finite-sum nonconvex optimization by proposing a stochastic gradient descent algorithm with nested variance reduction, which converges to an ε-approximate stationary point with a gradient complexity of Õ(n ∧ ε⁻² + ε⁻³ ∧ n¹/²ε⁻²), improving over prior methods like SVRG and SCSG.
We study finite-sum nonconvex optimization problems, where the objective function is an average of $n$ nonconvex functions. We propose a new stochastic gradient descent algorithm based on nested variance reduction. Compared with conventional stochastic variance reduced gradient (SVRG) algorithm that uses two reference points to construct a semi-stochastic gradient with diminishing variance in each iteration, our algorithm uses $K+1$ nested reference points to build a semi-stochastic gradient to further reduce its variance in each iteration. For smooth nonconvex functions, the proposed algorithm converges to an $ε$-approximate first-order stationary point (i.e., $\|\nabla F(\mathbf{x})\|_2\leq ε$) within $\tilde O(n\land ε^{-2}+ε^{-3}\land n^{1/2}ε^{-2})$ number of stochastic gradient evaluations. This improves the best known gradient complexity of SVRG $O(n+n^{2/3}ε^{-2})$ and that of SCSG $O(n\land ε^{-2}+ε^{-10/3}\land n^{2/3}ε^{-2})$. For gradient dominated functions, our algorithm also achieves better gradient complexity than the state-of-the-art algorithms. Thorough experimental results on different nonconvex optimization problems back up our theory.