Molei Tao’s First Research Paper on Machine Learning Takes Top Honors at AI Conference

This story was originally published by Georgia Tech Machine Learning.

A mathematician by trade, Molei Tao typically uses mathematics to design algorithms and solve physical science problems like how planets move. Recently, he became attracted to machine learning, an area that according to him, contains numerous interesting problems that are mathematically exciting and can benefit from modern mathematical tools.

This year, Tao, an associate professor in the School of Mathematics, published his first machine learning conference paper, and this work was awarded the best paper award at the 23^rd International Conference on Artificial Intelligence and Statistics (AISTATS).

His paper, Variational Optimization on Lie Groups with Examples of Leading (Generalized) Eigenvalue Problems, details a natural way for adding momentum to the gradient descent optimization in non-flat spaces. In flat spaces, the approach of adding momentum for accelerating the training of machine learning models has already been tremendously successful, and this new progress expands the applicability of the popular and powerful idea.

Tao felt fortunate to win this recognition. He and his co-author, Tomoki Ohsawa of the University of Texas at Dallas, had read many classical works from previous proceedings of AISTATS. Impressed with the quality of work, the authors chose to submit their first draft to it.

“We really did not think of winning the award at all. The completion of our work was around the AISTATS submission deadline, so we just submitted happily,” said Tao. “This submission confirmed to me how vibrant the machine learning community is. They are open to new ideas and many people made real efforts to understand this theoretical work and ask good questions.”

Tao was also encouraged by the collaborative and interdisciplinary environment provided by the Machine Learning Center at Georgia Tech (ML@GT) where Tao is also a faculty member.

“Georgia Tech is full of leaders in machine learning with different areas of expertise. We are proud of how Molei continues to innovate and further connect machine learning to real-world problems, both physically and computationally, and look forward to his future accomplishments” said Irfan Essa, ML@GT executive director.

Tao stated that winning this award encouraged him, and hopefully other mathematicians and scientists, to continue searching for fusion of ideas and creating new venues of applications.