Optimization and Learning in FACTORIE

We present the optimization package in the FACTORIE library for machine learning, graphical models, and natural language processing in Scala. The library is designed with an eye towards the way practitioners naturally break down the structure of a learning problem, giving first-class status to, for example, training examples, update rules, and regularization. It supports flexible mixing and matching of parameter representation, update rules, loss functions, online-vs-batch training schedules, and serial-vs-parallel processing.