Learning Directly From Data Rather Than Human-Provided Expertise

Alumnus Yong Sheng Soh (PhD ’18, ACM), advised by Professor Venkat Chandrasekaran, has won the 2018 INFORMS Optimization Society Student Prize for his paper entitled Learning Semidefinite-Representable Regularizers. Regularization-based algorithms are widely used in the solution of data-driven inverse problems arising in statistics, operations research, signal processing, and machine learning. These approaches typically require a great deal of prior domain-specific expertise about the structure of the underlying problem. Such expertise is especially challenging to obtain in the modern era of 'Big Data' in which our ability to sense all manner of phenomena and produce massive datasets far exceeds our ability to understand the underlying mechanisms governing the phenomena. In his award-winning work, which has been accepted for publication at the Journal of the Foundations of Computational Mathematics, Yong Sheng has developed a new framework for learning suitable regularization methods directly from data rather from human-provided expertise. [INFORMS Optimization Society Award] [Read the paper]