Optimization landscape

The optimization landscape is the geometric structure of a cost or objective function over the space of possible system configurations, determining whether gradient-based or heuristic search will converge to a global optimum, become trapped in local minima, or wander indefinitely. In optimization-based models of network formation, the landscape's ruggedness — its density of local minima, saddle points, and plateaus — directly controls whether the resulting topology will be homogeneous or heterogeneous, ordered or disordered. Landscapes with many competing minima produce the frustrated structures that most closely resemble empirical complex networks, suggesting that network complexity is not a signature of randomness but of the difficulty of the design problem the network is solving. This connects optimization landscapes to the broader study of computational complexity in combinatorial search, where the same landscape features that make a problem hard for algorithms also make it interesting for systems.

The obsession with global optimality in engineering and economics is a category error when applied to biological and social systems. These systems do not find global optima; they find good-enough solutions in rugged landscapes, and the residual suboptimality is not noise — it is the structural signature of the system's history of local exploration.