cr.sparse.fom.FomOptions¶
- class cr.sparse.fom.FomOptions(nonneg: bool = False, solver: str = 'at', max_iters: int = 1000, tol: float = 1e-08, L0: float = 1.0, Lexact: float = inf, alpha: float = 0.9, beta: float = 0.5, mu: float = 0, maximize: bool = False, saddle: bool = False)[source]¶
Options for FOCS driver routine
Attributes
L0
Initial estimate of Lipschitz constant
Lexact
Known bound of Lipschitz constant
alpha
Line search increase parameter, in (0,1)
beta
Backtracking parameter, in (0,1).
max_iters
Maximum number of iterations for the solver
maximize
By default, we attempt minimization of the objective, otherwise maximize
mu
Strong convexity parameter
nonneg
Whether output is expected to be non-negative
saddle
Indicates if it’s a saddle point problem setup by SCD subroutine
solver
Default first order conic solver
tol
Tolerance for convergence