What is the role of optimization tolerance in SVM training?

Optimization tolerance controls when SVM training stops based on convergence precision.

Tolerance defines how close the solution must be to optimal before stopping. Lower tolerance increases training time but improves accuracy slightly. It is important in large-scale optimization where exact convergence is expensive.