Combining data-drivenness and robustness in machine learning

Machine learning and optimization are two key technologies driving analytics in industry. The increasing availability of large datasets has motivated the design of new machine learning and optimization models that can better exploit data. 

Telfer professor Jonathan Li has received a Natural Sciences and Engineering Research Council of Canada (NSERC) Discovery grant for a project titled “An analytic framework for the simultaneous pursuit of data drivenness and robustness in machine learning and optimization.” 

Central to machine learning and optimization models are the principles of data-drivenness and robustness. The former refers to capturing information and the latter to protection against sampling errors and unforeseen data.

Li wishes to establish a data-driven distributionally robust optimization (DD-DRO) framework that can resolve, or at least lessen, the trade-off between data-drivenness and robustness.

Expected contribution to knowledge

Distributionally robust optimization (DRO) has emerged in the past decade as arguably one of the most high-impact methodologies across many fields of study, including statistics, computer science, finance and operations research. The increasing availability of large datasets has stimulated many studies on how to best incorporate data into DRO. 

Li’s research seeks to enhance DRO applicability and adoption in many fields. It could offer financial institutions a means to capitalize on the power of analytics in portfolio management, to build more efficient, resilient and equitable portfolios, furthering economic stability and offering financial resources to the disadvantaged.