Grad Student Xin Sun Assists Physics Team with Photonic Tech Project to Aid Artificial Neural Networks
Xin Sun , as a 5th year Ph.D. student in mathematics, was part of a team of researchers at MIT and elsewhere that has developed a new approach to such computations, using light instead of electricity, which they say could vastly improve the speed and efficiency of certain deep learning computations. Their results appeared recently in the journal Nature Photonics . “Deep Learning” computer systems are based on artificial neural networks that mimic the way the brain learns from an accumulation of examples. These systems enable technologies such as face- and voice-recognition software, and could search through medical data to find patterns for diagnostic use, or scan chemical formulas for possible new pharmaceuticals.
Sun, whose advisor was Scott Sheffield, studies probability theory and mathematical physics; he has done research on random planar geometry, including SLE, Gaussian free field, random planar maps and Liouville quantum gravity. He was recruited to this photonics project by his friend, physics post-doc Yichen Shen, one of the first two authors of the article, and part of a team scattered around the world.
“The team in the physics department was trying to use optical device to realize mathematical operations involved in the deep learning algorithm, which are matrix multiplication and nonlinear function,” he explained. “One problem they encountered is to represent every matrix by 2 by 2 unitary matrices with other simple operations that can be easily realized by an optical device.” Together with Yichen Shen , he found what he explained "an economic way with respect to the optical device."
A recent graduate, Xin will go in July to Columbia University as a Simons Junior Fellow He received his B.S. in mathematics from Peking University.
Congratulations Xin Sun!