SGDA-B is the first backtracking-enabled stochastic GDA algorithm for nonconvex-concave minimax problems that achieves the best known complexity bounds among methods agnostic to L, μ, and σ².
Basically, for given ϵ > 0, (xϵ, yϵ) is ϵ-stationary (in the sense of the (M4) metric) if ∃u ∈ ∇ xf(xϵ, yϵ) + ∂g(xϵ), ∃v ∈ −∇ yf(xϵ, yϵ) + ∂h(yϵ) : max {∥u∥, ∥v∥} ≤ ϵ
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
math.OC 1years
2024 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
A Stochastic GDA Method With Backtracking For Solving Nonconvex Concave Minimax Problems
SGDA-B is the first backtracking-enabled stochastic GDA algorithm for nonconvex-concave minimax problems that achieves the best known complexity bounds among methods agnostic to L, μ, and σ².