Tight Stochastic Condition-Number Dependence in Nonconvex-Strongly-Concave Minimax Optimization
En palabras de los autores
We study whether the linear condition-number dependence in the stochastic complexity of SAPD+ is necessary for nonconvex-strongly-concave minimax optimization. For jointly -smooth objectives with dual strong-concavity parameter , we prove a lower bound that matches the SAPD+ upper bound under the same Moreau-envelope stationarity criterion and the same primal-dual initialization gap. Specifically, when , the worst-case complexity of zero-respecting algorithms is in the stated accuracy regime, where , bounds the initial primal-dual gap, and bounds the variance of a general unbiased first-order oracle. The lower bound is realized on a smooth problem class with a bounded dual box. Our construction routes each link of a nonconvex zero-chain through a dual gradient of magnitude proportional to , while an undiscovered primal coordinate prevents stationarity. It also yields the primal-gradient lower bound after combination with the known deterministic bound, where bounds the initial primal function gap.
Apareció: lunes, 28 de septiembre. arXiv. Preprint, todavía sin revisión por pares.
Comentario de los autores: 20 pages