Solves convex cone programs via operator splitting. Can solve: linear programs ('LPs'), second-order cone programs ('SOCPs'), semidefinite programs ('SDPs'), exponential cone programs ('ECPs'), and power cone programs ('PCPs'), or problems with any combination of those cones. 'SCS' uses 'AMD' (a set of routines for permuting sparse matrices prior to factorization) and 'LDL' (a sparse 'LDL' factorization and solve package) from 'SuiteSparse' (< https://people.engr.tamu.edu/davis/suitesparse.html>).
| Parameter | Default Value | Type | Description |
|---|---|---|---|
| max_iters | 100000 | integer | maximum number of iterations |
| eps_rel | 1e-04 | numeric | relative feasibility tolerance |
| eps_abs | 1e-04 | numeric | absolute feasibility tolerance |
| eps_infeas | 1e-07 | numeric | Infeasibility tolerance (primal and dual) |
| alpha | 1.5 | numeric | Douglas-Rachford relaxation parameter |
| rho_x | 1e-06 | numeric | Primal scale factor |
| scale | 0.1 | numeric | Initial dual scale factor, updated if adaptive_scale is True |
| verbose | FALSE | boolean | Write out progress |
| normalize | TRUE | boolean | Whether to perform heuristic data rescaling |
| warm_start | FALSE | boolean | Whether to warm start |
| acceleration_lookback | 0 | integer | How much memory to use for Anderson acceleration |
| acceleration_interval | 1 | integer | Run Anderson acceleration every this number of iterations |
| adaptive_scale | TRUE | boolean | Whether to heuristically adapt dual scale through the solve |
| log_csv_filename | string | If non NULL SCS will write logs of quantities through the solver (ignored) |
|
| time_limit_secs | 0 | numeric | Time limit for solve run in seconds (can be fractional, 0 implies no limit |
| write_data_filename | string | If non NULLset the problem data is dumped to this filename (ignored) |
https://doi.org/10.1007/s10957-016-0892-3.