Zeroth-Order Optimization Finds Flat Minima

Bingcong Li (ETH Zurich) · Liang Zhang (Renmin University of China) · Michael Muehlebach (Max-Planck Institute) · Niao He (ETH Zurich) · Sewoong Oh (University of Washington) · Kiran Thekumparampil (Amazon)

binary classification black-box attacks convergence rates convex functions convex lossesflat minimaimplicit regularizationlanguage model fine-tuningoptimization theory reinforcement learningsharp minimasmooth functions stationary pointtrace of hessianzeroth-order methods

Zeroth-order methods are extensively used in machine learning applications where gradients are infeasible or expensive to compute, such as black-box attacks, reinforcement learning, and language model fine-tuning. Existing optimization theory focuses on convergence to an arbitrary stationary point, but less is known on the implicit regularization that provides a fine-grained characterization on which particular solutions are finally reached. We show that zeroth-order optimization with the standard two-point estimator favors solutions with small trace of Hessian, which is widely used in previous work to distinguish between sharp and flat minima. We further provide convergence rates of zeroth-order optimization to approximate flat minima for convex and sufficiently smooth functions, where flat minima are defined as the minimizers that achieve the smallest trace of Hessian among all optimal solutions. Experiments on binary classification tasks with convex losses and language model fine-tuning support our theoretical findings.