This project is inspired by the phenomenon of conical refraction discovered by Hamilton in the 1830’s, in which a single ray of light refracts into a cone upon entering a bi-axial crystal. I am interested in studying similar phenomena for more general equations of physical interest; like those of elasticity theory. My approach to this problem is based on the use of topological methods to predict the existence of singularities in the Fresnel Hyper-surface associated to hyperbolic partial differential equations.
I am interested in the study diffusion processes taking place inside complex geometrical objects. More specifically, I study of diffusion processes on thin channels (e.g. capillary veins, nano tubes, etc) using tools from differential geometry and topology. For channel-like object the multi-dimensional diffusion equation can be “reduced” to a diffusion equation in a 1-dimensional spatial variable. The diffusion coefficient for this reduced equation is known as the effective diffusion coefficient. My current goal is in find an analytic formula for this coefficient for channels in 3-dimensional space with cross sections of arbitrary shape.
I am trying to understand diverse geometrical aspects of neural networks by applying ideas used in computer graphics to construct geometrical objects. I use a technique to smoothly blend regions in space (using smooth set operations like intersections, unions and differences) to motivate the use of higher order neural networks (e.g. quadratic and sigma-pi neural networks).