You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I'm trying to use scs to solve a collection of synthesis convex qp problems. The parameters of the instances are randomly generated. Though, all the problems are successfully solved, the number of reported iterations are all taking values at 75, 50 things like that. For a specific problem, I tried to warm start the solver from different point, however, the number of iteration is still unchanged. Controlling the tolerance for the residuals also would not help. Is there any specific settings that I need to turn off to get the real number of iterations for the solver to converge? Thank you~
The text was updated successfully, but these errors were encountered:
The convergence criteria are only checked every 25 iterations by default in order to save on compute time (since checking the criteria is about the same computational cost as taking a step in the algorithm). I would tighten the tolerances, or try generating hard problems first. If you want to change the 25 parameter to another number, you need to change this line in the code and then recompile the solver.
I'm trying to use scs to solve a collection of synthesis convex qp problems. The parameters of the instances are randomly generated. Though, all the problems are successfully solved, the number of reported iterations are all taking values at 75, 50 things like that. For a specific problem, I tried to warm start the solver from different point, however, the number of iteration is still unchanged. Controlling the tolerance for the residuals also would not help. Is there any specific settings that I need to turn off to get the real number of iterations for the solver to converge? Thank you~
The text was updated successfully, but these errors were encountered: