I will present and discuss a number of computational experiments associated
with the coarse-graining of atomistic/agent-based simulations. In particular, I
will discuss coarse reverse integration, as well as the use of diffusion maps
(a manifold-learning technique) to suggest the selection of certain coarse-grained observables
(reduction coordinates) for the atomistic simulations. The illustrations will come
from molecular dynamics, Monte Carlo as well as agent-based models.