Towards new global optimization paradigms

With ever increasing computational power available, tasks that were once impractical to compute can now be done routinely. Many methods to resolve complex designs rely on designing smaller subsystems, making simplifications or assumptions of their interactions. Separating the problem into subsystems makes local optimization easier through reduced dimensionality. However, considering components independently can miss globally optimal solutions only found in their interactions. To account for all conceivable end use cases, subsystem solutions may be overly constrained and conservative when integrated in the complete system. Growth in computational power makes expanding the scope of optimization realistic. Through physical examples, this work demonstrates the performance gains possible by expanding optimization to include adjacent sub systems. Presented results find improvements in optimization efficiency and effectiveness against published methods, metrics rarely compared despite being essential for practical use. Quantization of variables, to include real-world constraints, is demonstrated without appreciably reducing optimization efficiency. Lossless flattening of a problems dimensions into a non-dimensional graph is shown, mapping numerical problems to evolutionary algorithms. This mapping generalizes distance based search methods onto new problems. Application of these new tools for optimization of electronic circuit topology finds no known methods are significantly more efficient than pure chance. Exploring optimization algorithms, little improvement in optimization efficiency is found over existing known methods. However, major gains in computational efficiency are shown by using appropriate simulation methods. New, more complete, models are then demonstrated to be computed in less time than existing methods while also returning more accurate results. This produces: faster; more robust; and higher performing solutions than prior works. PhD by publication