Near-Optimal Non-Convex Stochastic Optimization under Generalized Smoothness
This work addresses computational inefficiencies and lack of single-run guarantees in non-convex optimization for machine learning applications, though it is incremental as it builds on existing STORM variants.
The paper tackles the problem of stochastic optimization under generalized smoothness, achieving near-optimal high-probability sample complexity with constant batch size, specifically O(log(1/(δε))ε^{-3}) for an O(ε)-stationary point.
The generalized smooth condition, $(L_{0},L_{1})$-smoothness, has triggered people's interest since it is more realistic in many optimization problems shown by both empirical and theoretical evidence. Two recent works established the $O(ε^{-3})$ sample complexity to obtain an $O(ε)$-stationary point. However, both require a large batch size on the order of $\mathrm{ploy}(ε^{-1})$, which is not only computationally burdensome but also unsuitable for streaming applications. Additionally, these existing convergence bounds are established only for the expected rate, which is inadequate as they do not supply a useful performance guarantee on a single run. In this work, we solve the prior two problems simultaneously by revisiting a simple variant of the STORM algorithm. Specifically, under the $(L_{0},L_{1})$-smoothness and affine-type noises, we establish the first near-optimal $O(\log(1/(δε))ε^{-3})$ high-probability sample complexity where $δ\in(0,1)$ is the failure probability. Besides, for the same algorithm, we also recover the optimal $O(ε^{-3})$ sample complexity for the expected convergence with improved dependence on the problem-dependent parameter. More importantly, our convergence results only require a constant batch size in contrast to the previous works.