### Data: Yearly NIPS collaboration networks

The following data contains the yearly collaboration networks of authors who published in the NIPS

conference in the years 1987-2012 (26 years).

If authors *i* and *j* co-authored a paper which appeared at year *t*, then the network at time *t*

will contain an edge between *i* and *j*. If a set of authors co-authored a paper, then the network

would contain an edge between each pair within that set.

The data includes a total of 5,722 authors, 4,798 papers, and 13,608 co-authorship diads.

The data was parsed from the NIPS proceedings records published on DBLP.

The raw text can be found here.

Name dissambigutaions have been resolved manually. The file authors.csv contains a mapping

between names appearing on the paper and the node id in the network.

The data is in MATLAB format. It consists of a cell array 'yearly_collabs' with 26 entries, each for a

single year. Each entry holds a 5722x5722 sparse (symmetric) matrix which represents the collaboration

network for the corresponding year.

DOWNLOAD DATA

If you use this data in published works, please cite us with this source:

Learning Structured Models with the AUC Loss and Its Generalizations.

Rosenfeld, N., Meshi, O., Tarlow, D., & Globerson, A. (2014).

In Proceedings of the Seventeenth International Conference on Artificial Intelligence and Statistics (pp. 841-849).