Abstract
dc:descriptionEvery graph G can be associated with many well-known invariant properties along with their corresponding values. A framework is proposed to measure the change in any particular invariant upon addition of a new edge $e$ in the resulting graph G+e. In graphs, the P-impact of an edge e is the `magnitude' of the difference between the values of the invariant P in graphs G+e from G. Several famous invariants are explored and a proof towards optimal edge addition for distance-impact in trees is given. A natural application to measuring the impact of edge addition to a graph is that of link prediction. An efficient algorithm for link prediction even with cold-start vertices using a subspace sharing method that decouples matrix completion and side information transduction is presented. This method is extended to predict ratings in user-item recommender systems where both may be cold-start.
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Ross, Dennis
- Contributors dc:contributor
-
- Esfahanian, Abdol-Hossein
- Tan, Pang-Ning
- Xing, Guoliang
- Nejadhashemi, Amir Pouyan
Subjects
dc:subject × 3Rights
dc:rights- Statement dc:rights
-
- In Copyright
- Language dc:language
- English
Identifiers
dc:identifier.*- Identifier
-
etd:3985
isbn:9781339828657
isbn:1339828650
oclc:961288566
umi:10124766
local:Ross_grad.msu_0128D_14558 - OAI record oai:identifier
- https://d.lib.msu.edu/oai?verb=GetRecord&metadataPrefix=oai_dc&identifier=oai%3Ad.lib.msu.edu%3Aetd_3985