ECE professor serves machine learning theory community as co-chair of premier conference
Working at the intersection of control theory and machine learning, ECE Associate Professor Max Raginsky explores fundamental questions about how to design complex systems that automatically learn to perform various tasks better on the basis of experience despite changing requirements and environments. His research results have applications in electronic circuit design automation, autonomous systems, and artificial intelligence.
This past year, Maxim Raginsky was appointed co-chair of the premier international conference on machine learning theory and artificial intelligence. The 35th annual Conference on Learning Theory (COLT) is scheduled to be held July 2-5, 2022, in London.
“I view this as a huge vote of confidence by the [learning theory] community,” said Raginsky, the William Everitt Fellow in ECE and a faculty member in the Coordinated Science Lab. “It’s a big responsibility and I want to do my part to keep the conference intellectually vibrant and socially relevant.”
“The selection of Professor Raginsky as a co-chair of the world’s premiere conference on the theory of machine learning is a strong testament to the breadth and depth of his work in the field,” said ECE Department Head Bruce Hajek. “It is wonderful to have him among our ranks.”
Over the past decade, Raginsky has contributed to COLT as both a research presenter and reviewer. He is also an elected member of the Association for Computational Learning Board of Directors that oversees the conference. Despite the huge time commitment involved in chairing a conference, Raginsky was honored to do it.
“Service is important and I want to make sure I contribute to the mentoring and growth of knowledge,” Raginsky said, noting that he shares organizing duties with co-chair Po-Ling Loh, a faculty member at the University of Cambridge in England.
According to Raginsky, COLT is unique because it’s a small conference—roughly 150 papers are presented every year—that focuses on the theory underpinning machine learning.
“What unites the [COLT] community at this conference is the emphasis on mathematical precision,” he said. “COLT is where we talk about what can or cannot be done with certain machine learning systems.”
Machine learning systems include recommendation algorithms used by retail and entertainment companies, automatic text reply suggestions used by email providers, and smart thermostats that collect large amounts of usage data and make recommendations for saving energy.
“We state results in the form of mathematical theorems whose proofs can be verified and checked by others, so we understand what machine learning algorithms can do quantitatively,” he said.