Functional maps introduced to the computer-graphics community in 2012 by Bosanko et al are popular tools for expressing correspondences between shapes. These make use of linear algebra to represent a map from one surface into another. In this project, we will formulate and implement a kernelized version of functional maps, using ideas from kernel methods in machine learning. The hope is to improve the quality and reliability of correspondence algorithms.

I am super-excited to be working with Professor Justin Solomon on this SuperUROP. Because his computer graphics class piqued my interest in computer graphics, I can’t wait to delve deeper. Through this SuperUROP project, I hope to get a deeper understanding of mathematics and numerical algorithms as well as learning whether research is something I want to pursue.