Constrained Second-Order Minimax Optimization
Descriptions may be stale. Consult the linked Lean sources.
Constrained Second-Order Minimax Optimization
We consider minimizing
The optimization problem is
The parameters satisfy
Our assumptions:
Assumption 1 (Second-order regularity). The function
Assumption 2 (Strong concavity). For every
Assumption 3 (Initial bounds). The initial value gap
and full gradient satisfy
Queries may be made at any
Results
Finite-dimensional real inner-product spaces
| Result | Oracle complexity | Novelty | Limitations |
|---|---|---|---|
| (1) ConstrainedMinimaxCubic
(Ours) |
|||
| (2) Minimax-PPA (Lin et al. 2020, Theorem 20) | (1): Scope: unbounded
closed convex |
(1) needs full Hessians,
ambient Hessian Lipschitzness and |
|
| (3) (Zhang et al. 2026, Theorem 4.1, v2) | (1): Scope: closed convex
dual constraints, including changing active constraints. Matches leading
|
(1) has |
|
| (4) ECRN (Zhou et al. 2026, Theorem 11, v1) | (1): Scope: closed convex
dual constraints. Matches leading |
(1) guarantees only first-order stationarity; retains explicit initialization and tracking costs. |
Abbr: