Title: Hierarchical eigenmodels for pooling relational data


Abstract: A relational dataset may contain measurements of different types
or measurements under different conditions. For example, social network
data often consists of repeated measurements over time, and gene
expression data may be gathered on different types of patients. In both
cases, the data can be represented as multiple matrices. Variability
across such matrix-valued datasets can be described using a probability
model for matrix eigenstructure. Such a model can leverage any
across-matrix similarity in eigenstructure to improve inference relating
to any one matrix, and can provide a description of the main patterns
exhibited across matrices.  In this talk I present two examples of
eigenstructure modeling, and discuss the tools required for Bayesian
inference.