“…Chi and Kolda showed in [17] that under these assumptions a Poisson CP tensor model is an effective low-rank approximation of X. The Poisson CP tensor model has shown to be effective in analyzing latent patterns and relationships in count data across many application areas, including food production [13], network analysis [11,19], term-document analysis [16,29], email analysis [14], link prediction [18], geosptial analysis [22,28], web page analysis [39], and phenotyping from electronic health records [27,30,31] One numerical approach to fit low-rank Poisson CP tensor models to data, tensor maximum likelihood estimation, has proven to be effective. Computing a Poisson CP tensor model via tensor maximum likelihood estimation involves minimizing the following non-linear, nonconvex optimization problem:…”