Earlier this month, TRIAD, the Transdisciplinary Research Institute for Advancing Data Science at Georgia Tech, hosted a workshop on the Theoretical Foundation of Deep Learning for over 100 participants from both academia and industry.

Deep learning has been a major driving force in the recent surge of interest in artificial intelligence (AI) and machine learning. While deep learning has resulted in tremendous empirical success, the theoretical understanding of deep learning remains an important research field rife with opportunity.

“In other words, understanding why and how AI works or doesn’t work is not clear,” said Xiaoming Huo, a professor in the H. Milton Stewart School of Industrial and Systems Engineering (ISyE) and executive director of TRIAD.

“The theory needs significant development, so the workshop brought together people who have been working on this problem and enabled them to exchange ideas,” he noted.

The workshop’s interdisciplinary focus was reflected in both the diversity of speakers, including experts in fields ranging from statistics and optimization to computer science and applied mathematics, and its sponsors, which included ISyE; Georgia Tech’s colleges of Computing, Engineering, and Sciences; Hong Kong University of Science and Technology’s Big Data Institute; the Statistical and Applied Mathematical Sciences Institute (SAMSI); and the American Statistical Association (ASA).

Promising ideas on the theoretical foundation of deep learning have emerged, and the TRIAD workshop provided an avenue for researchers in related fields to review existing work, communicate new results, and seek new research directions.

About TRIAD

TRIAD is a cross-disciplinary institute that was established in 2017 as part of the National Science Foundation’s TRIPODS (Transdisciplinary Research in Principles of Data Science) program. TRIAD unites statistics, mathematics, and theoretical computer science to further develop the foundations of data science. TRIAD brings together senior, mid-career, and junior faculty members, postdoctoral fellows, graduate and undergraduate students, all from Tech’s colleges of Computing, Engineering, and Sciences, and data science practitioners-at-large using focused working groups, national and international workshops, and organized innovation labs.