Distances in Partial Orders for Knowledge Discovery
Date: 3:30pm - 4:30pm PST November 7, 2012 Location: Howard 254
In many disciplines it is helpful to have a mathematical structure in which to organize knowledge of the subject. Common structures for this purpose are partial orders and directed graphs which we use to quantify similarity and distance between sets of concepts. At PNNL, this is used within the Signature Discovery Initiative which has a mission to construct and detect signatures, asses signature quality, and adapt to dynamically changing signatures. The running hypothesis of the initiative is that the signature discovery process is generalizable across disciplines. In order for us to be able to generalize this process we must be able to map processes from one domain to another, e.g., from biological sequences to cyber sequences. In this talk I will describe how we use partial orders and graphs to organize knowledge and compare subject areas, give an overview of current distance and similarity measures, and finally discuss our methods for improving upon these distances and similarities.