Equation Solution Figures of Merit, Metaheuristic Search, and the Schrodinger Equation

This presentation deals with: a definition of ``equation error''; a consideration of equation solution figures of merit based on equation error, and on other measures; and the use of metaheuristic techniques in the search for approximate solutions. These considerations are illustrated by application to the Schrodinger equation for a simple system. Models suitable for computation are produced. Computation results are used to compare the consequences of selection of different figures of merit. ``Equation error'' is defined to be the quantity by which an approximate solution fails to satisfy an equation. ``Equation error variance'' is defined to be the squared modulus of the equation error summed/integrated over the domain of interest. (Generalization to sets of equations is straightforward.) In the example, equation error variance is a functional of the Schrodinger wave function. Possible figures of merit include: ground state energy, system geometry, and equation solution variance. The (derivative-free) metaheuristic used to solve the Schrodinger equation has been changed from a genetic algorithm, used in earlier versions of this research, to evolution strategy with covariance matrix adaptation.