Home > Events > Oral Candidacy - Joseph Crawford

Oral Candidacy - Joseph Crawford

Start: 5/5/2017 at 11:00AM
End: 5/5/2017 at 2:30PM
Location: 117 I Cushing Hall
Attendees: Faculty and students are welcome to attend the presentation portion of the defense.
Add to calendar:
iCal vCal

Joseph Crawford

Oral Candidacy

May 5, 2017        11:00 am        117I Cushing

Adviser:  Dr. Tijana Milenkovic

Committee Members:

Dr. Ron Metoyer        Dr. Aaron Striegel        Dr. Tim Weninger

 

Title:

“Discovering Important Patterns across Networks via Network Alignment and Clustering”

Abstract:

Networks (or graphs) can be used to model real-world phenomena in a variety of domains. For example, in the biological domain, they can be used to represent protein-protein interactions, and in the social domain, they can be used to represent interactions between people. There exist many computational strategies for network analysis, each answering a different applied question. Two of the most popular computation strategies are network alignment (NA) and network clustering (NC), as both strategies are able to extract important, common structures that are "hidden" between multiple networks. The discovery of important structures across networks is useful in many computational and applied problems, and thus, we aim to advance the fields of both NA and NC.

In terms of our past and present work, in the context of NA, we present a set of general recommendations for a fair evaluation of NA methods, as well as three new NA methods that improve upon existing methods due to their novel approaches to solving the NA problem. In the context of NC, we present a new method for clustering a dynamic network that addresses key drawbacks of existing methods for the same purpose.

In terms of the proposed work, first, as we have done with the field of dynamic NC, we propose a new method for heterogeneous NC that is expected to address key drawbacks of existing methods for the same purpose. Second, we propose to further improve the NA field by incorporating the idea of NC into the process of constructing an alignment between compared networks, unlike what traditional NA approaches do.