Incorporating Graph-Based Models in a Deep Learning Framework for Operational Face Recognition
We organize a collection of biometric records (gallery) into a graph structure to aid with face recognition, biographic prediction, and other tasks.
Auto-Tuned Models (ATM)
ATM attempts to save data scientists time by automating the classifier selection process. Typically, a data scientist will manually select a classifier and attempt to tune the hyperparameters. ATM accepts a feature set and tries a variety of methods as well as find the best hyperparemters.