Wesley Lao

My research aims to marry numerical methods with deep learning techniques in order to accelerate and augment expensive numerical solvers. By leveraging existing work in computational science, I hope to extend the capabilities of computational tools, rather than reinvent them. In particular, I am interested in using geometric deep learning (i.e., graph-based methods) to bring the advances in deep learning for computer vision and networks into the realm of physics-based problems. Our world and the phenomena that govern it are best represented with unstructured geometries capable of resolvng multiple scales effectively. I believe baking this underlying principle into the design of deep learning models, rather than forcing the physical domain to assimilate to an ill-suited format, will be instrumental in training accurate and reliable models.

Interests

Education