What can be done to attain near-global optimization in complex systems?

[oscormdigital20]

The target optimization functional can often be non-convex, with numerous local extrema. Overall, it is hard to locate a global, or almost global optimum. But it is common to run general coarse grained models of flow networks with local positive feedback rules on the conductivity of the vessels into low efficiency local minima. Oscorm demonstrates the development of the underlying issues in Global Cost Optimization, related to the dynamical equations of network development.