Automating data augmentation by learning a generative sequence model over user-specified transformation functions.

Proposing a structure estimation method that is 100x faster than a maximum likelihood approach for training data.

Presenting Coral, a paradigm that infers generative model structure, significantly reducing the amount of data required to learn structure.

A paradigm for labeling training datasets programmatically rather than by hand.

Introducing DDLite, an interactive development framework for data programming.