Negating duals in Hessian

[tso-martin]

Dear @mkhraijah,
I wonder why the duals of all constraints are negated in https://github.com/mkhraijah/PowerModelsADA.jl/blob/main/src/algorithms/aladin_coordinated_methods.jl#L403. Is there a special reason for this?
Best regards,
Martin

[mkhraijah]

Hi @tso-martin,
This implementation follows the heuristic in Section A3 of the paper: ALADIN-$\alpha$—An open-source MATLAB toolbox for distributed non-convex optimization. We negate the dual variables to account for the JuMP convention on the sign of dual variables. We also negate any negative eigenvalue to ensure the Hessian matrix is positive-semidefinite. In any case, we are working on rewriting the ALADIN code using the new nonlinear features in JuMP. So, this part will be changed.
Regards,
Mohannad