Mitchell Centre seminar
| Dates: | 19 April 2023 |
|---|
| Times: | 16:00 - 17:00 |
|---|
| What is it: | Seminar |
|---|
| Organiser: | School of Social Sciences |
|---|
|
|
Jürgen Pfeffer
Technical University of Munich
CONCOR Revisited: Algebraic Clarifications and Practical Implications
In 1975, Breiger et al. presented an algorithm, CONCOR (an acronym derived from " convergence of iterated correlations"), for bipartite partitioning of network data. While most clustering algorithms aim to identify cohesive subgroups of nodes that are more densely connected among themselves than to other nodes in the network, the aim of the CONCOR algorithm is to obtain a partition of the vertices of the network graph into two blocks, in which the vertices in each block are considered to be structurally equivalent, or at least structurally similar. CONCOR is very versatile and can be applied to multi-relational networks and can even incorporate network attributes into the clustering procedure. However, the algorithm was criticized in the past for having opaque arithmetic properties. The major drawback of CONCOR can be summarized with Arabie & Schleutermann's (1990) observation that "CONCOR is a convenient algorithm which happens to yield substantive results rather than a compelling model as its mathematical properties are still not fully understood." The goal of this paper is to create a better understanding of algebraic and algorithmic properties of the CONCOR algorithm and to offer new perspectives on possible application scenarios.
Travel and Contact Information
Find event
G7
Humanities Bridgeford Street
Manchester
