Stochastic Bilevel Optimization on Nonconvex Stochastic Optimization
-3Sample ComplexityTTSA
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| TTSA(UL) f=C_L^{1,1}, (LL) g=SC and C_L^{2,2}, Noise Assumption=Bounded Variance, Hessian Inversion=Neumann approx., Single-Loop=Yes2026.06 | -3 | — | — | |
| BiSLS(UL) f=C_L^{1,1}, (LL) g=SC and C_L^{2,2}, Noise Assumption=Bounded Variance, Hessian Inversion=SGD, Single-Loop=Yes2026.06 | -3 | — | — | |
| AccBO(UL) f=Unbounded Smooth, (LL) g=SC and C_L^{2,2}, Noise Assumption=Small Variance O(ϵ), Hessian Inversion=SGD, Single-Loop=Yes2026.06 | -3 | — | — | |
| MA-SOBA(UL) f=C_L^{1,1}, (LL) g=SC and C_L^{2,2}, Noise Assumption=Bounded Variance, Hessian Inversion=SGD, Single-Loop=Yes2026.06 | -2 | — | — | |
| RQ-TTSA(UL) f=C_L^{1,1}, (LL) g=SC and C_L^{2,2}, Noise Assumption=Heavy-tailed (p ∈ (1, 2]), Hessian Inversion=SGD, Single-Loop=Yes2026.06 | 3 | — | — |