Predicting Program Properties from "Big Code"
Abstract
We present a new approach for predicting program properties from massive codebases (aka "Big Code"). Our approach first learns a probabilistic model from existing data and then uses this model to predict properties of new, unseen programs.