Conditional Expectations and Renormalization

Alexandre J. Chorin

Published 2002-04-17Version 1

In optimal prediction methods one estimates the future behavior of underresolved systems by solving reduced systems of equations for expectations conditioned by partial data; renormalization group methods reduce the number of variables in complex systems through integration of unwanted scales. We establish the relation between these methods for systems in thermal equilibrium, and use this relation to find renormalization parameter flows and the coefficients in reduced systems by expanding conditional expectations in series and evaluating the coefficients by Monte-Carlo. We illustrate the construction by finding parameter flows for simple spin systems and then using the renormalized (=reduced) systems to calculate the critical temperature and the magnetization.