Model Management Systems: Machine learning meets Database Systems

This project aspires to develop a computational framework and associated techniques towards model management systems, i.e., systems that extend traditional DBMS functionality to allow users to focus on specifying predictive tasks with limited direct involvement in lower-level decisions for the management of data, models, and computational resources. The prototype system to be developed in this project is implemented as a software layer between the DBMS and the user interface. Internally, the system implements techniques for model-based views and indexes for predictive queries, lazy incremental learning, and human-in-the-loop data and model curation.