Abstract:
|
In recent years, multiscale models have become popular for modeling spatiotemporal phenomenon. One advantage of these models is that the computation is scalable for a given multiscale partition. The specified partition controls the spatial structure in the model. In many applications there may be problem specific reasons to choose a particular multiscale partition; however, there may be more appropriate spatial partitions for a given problem. We propose a prior that enables a mixture of multiscale partitions or a procedure for choosing the maximum a posteriori partition.
|