Please checkout our first flagship product Http://HoodPantry.com.
Deep Learning, shown great promises in jointly learning features representation with final task-oriented optimizations to provide the best outcomes given the data, whether its is labeled or unlabeled. By harnessing massively parallel processing systems and high-end GPUs we train deep networks that are tailored towards complex tasks and decision making situations.
Probabilistic Graphical Models, when people gave up on Lisp and Prolog as they didn't address the demands of probabilistic inference, PGMs were introduced that addresses the problem both from a scalability point of view and soundness point of view. By updating a priori probabilities and weights as new data comes in PGMs match relational relationships of the data better and better.