Applying row-column permutation to matrix representations of large citation networks

摘要

A combinatorial optimization technique was used to rearrange a matrix of citations among 16,432 neuroscientists in which cited authors were columns, citing authors were rows, and an element was a count of citations of a cited author by a citing author. Optimization consisted of permuting rows and columns to maximize density near the diagonal. A scatterplot-type display of the permuted matrix revealed both global structure and local patterns of communication within neuroscience. The algorithm is described and compared to two other methods of mapping large networks, cluster analysis and correspondence analysis.