Stochastic Minimax Optimization on Theoretical Analysis
-4Sample ComplexitySGDA
Evaluation results
| Method | Links | |
|---|---|---|
| SGDACost Structure=Nonconvex–SC, Lipschitz Condition=Lipschitz risk gradient, Learning Rates=Constant2024.06 | -4 | |
| Stoc-Smoothed-AGDACost Structure=Nonconvex–PL, Lipschitz Condition=Lipschitz risk gradient, Learning Rates=Constant2024.06 | -4 | |
| SREDACost Structure=Nonconvex-SC, Lipschitz Condition=Expected Lipschitz loss gradient, Learning Rates=Constant2024.06 | -3 | |
| DM-HSGDCost Structure=Nonconvex-SC, Lipschitz Condition=Lipschitz loss gradient, Learning Rates=Constant2024.06 | -3 | |
| MSGDACost Structure=Nonconvex–PL, Lipschitz Condition=Lipschitz loss gradient, Learning Rates=Diminishing2024.06 | -3 | |
| HCMM-1Cost Structure=Nonconvex–PL, Lipschitz Condition=Lipschitz risk gradient and Hessian, Learning Rates=Nonlinear via clipping2024.06 | -3 | |
| HCMM-2Cost Structure=Nonconvex–PL, Lipschitz Condition=Lipschitz risk gradient and Hessian, Learning Rates=Nonlinear via normalization2024.06 | -3 | |
| Extra-MomentumCost Structure=Strongly Monotone, Lipschitz Condition=Lipschitz risk gradient, Learning Rates=Constant2024.06 | 1 | |
| VR-AGDACost Structure=PL-PL, Lipschitz Condition=Lipschitz loss gradient, Learning Rates=Constant2024.06 | 9 |