An introduction to the Partition Decoupling Method (PDM) and an application to the U.S. equities market

Scott Pauls

Mathematics Department, Dartmouth College


Greg Leibon, Dan Rockmore, Rob Savell and I recently developed a new statistical learning method which we call the Partition Decoupling Method that, in contrast to established hierarchical methods, is aimed at decomposing complex systems into multiple overlapping partitions of the system. In this talk, I will introduce the method via an initial application to the U.S. equities market where the partitions reveal interesting geometric and functional structure. (Attached is a copy of our first paper on the subject).

Back to ACMS schedule