Lin, Juan K. (2003), Latent Variable Models for Link Analysis of Similarity Data, Computing Science and Statistics, 35, I2003Proceedings/LinJuan/LinJuan.paper.pdf ,
There is a need for statistical models which can organize large collections of pair-wise similarity relationships between objects into meaningful clusters. Similarity data consists of non-negative quantitative measurements of similarity between object pairs. Examples include internet connectivity data and document word count data. We present various latent variable models for finding reduced rank structure in similarity data. Applications will be presented in unsupervised clustering, targeted clustering based on pre-defined cluster relationships, and graph layout using reduced rank graph approximations.