John Straub
My lecture will present novel computational methods designed to solve challenging global optimization problems and to provide enhancement in the rate of conformational sampling in simulations of complex systems such as liquids and biomolecules.
A novel method for enhanced sampling, a generalization of the parallel tempering technique, will be presented. The method involves running several parallel and independent simulations, each parameterized differently within a generalized thermostatistics, where configurational exchanges are allowed among the independent trajectories. The algorithm is shown to be more effective than standard parallel tempering methods.
A variational method for finding optimal annealing schedules will be presented. The method is suitable for simulated annealing methods in which more than a single parameter is adjusted, such as quantum mechanical annealing where the annealing occurs through a variation of both temperature and Planck's constant. To demonstrate the technique, optimal cooling schedules are determined both for simulated annealing methods involving path integrals and within a generalized thermostatistics.
For each method, pedagogically useful examples will be presented to demonstrate the advantages of the technique relative to pre-existing methods.