Never Go Full Batch (in Stochastic Convex Optimization)

We study the generalization performance of full-batch optimization algorithms for stochastic convex optimization: these are first-order methods that only access the exact gradient of the empirical risk (rather than gradients with respect to individual data points), that include a wide range of algorithms such as gradient descent, mirror descent, and their regularized and/or accelerated variants. We provide a new separation result showing that, while algorithms such as stochastic gradient descent can generalize and optimize the population risk to within ε after (1/ε²) iterations, full-batch methods either need at least Ω(1/ε⁴) iterations or exhibit a dimension-dependent sample complexity.