Establish whether the particle-swarm optimization reaches the global loss minimum
Determine whether the parameters obtained from the single finite-budget particle-swarm optimization run represent the global minimum of the machine-learning correction’s training loss.
References
Thus, the optimization produces a substantially larger relative reduction in the loss for B3LYP than for r$2$SCAN. This difference is consistent with the smaller initial training-set errors of r$2$SCAN, although the single optimization run does not allow us to determine whether the resulting parameters represent the global minimum of the loss function.
— What Does a Semilocal Machine-Learning Correction Actually Learn? Size-Dependent Errors across Four Parent Functionals
(2609.37571 - Bhattacharjee et al., 29 Sep 2026) in Section 3, subsection “Training”
Multi-start refinement improves practical repeatability but does not prove global convergence or parameter uniqueness.
— When voltage sensors fail: Electrochemically constrained fault-tolerant state estimation for flat-plateau LFP batteries
(2610.05906 - Guo et al., 5 Oct 2026) in Appendix A, Section “Bound-constrained multi-start optimization”