Prof. Paul G. Mezey Canada Research Chair in Scientific Modeling and Simulation Department of Chemistry and Department of Physics and Physical Oceanography Memorial University of Newfoundland Canada E-mail: email@example.com
His Keynote Lecture was: Multidimensional Optimization Methods with Fewer Steps Than the Dimension: A Case of "Insider Trading" in Chemical Physics
"Insider trading" in commerce takes advantage of information that is not commonly available, and a somewhat similar advantage plays a role in some specific, very high-dimensional optimization problems of chemical physics, in particular, molecular quantum mechanics. Using a specific application of the Variational Theorem for the expectation value of molecular Hamiltonians, an optimization problem of thousands of unknowns does often converge in fewer than hundred steps. The search for optimum, however, is typically starting from highly specific initial choices for the values of these unknowns, where the conditions imposed by physics, not formally included in the optimization algorithms, are taken into account in an implicit way. This rapid convergence also provides compatible choices for "hybrid optimization strategies", such as those applied in macromolecular quantum chemistry . The efficiency of these approaches, although highly specific for the given problems, nevertheless, provides motivation for a similar, implicit use of side conditions for a better choice of approximate initial values of the unknowns to be determined. . P.G. Mezey, "On the Inherited "Purity" of Certain Extrapolated Density Matrices", Computational and Theoretical Chemistry, 1003, 130-133 (2013).