Using mathematics to untangle networks

I will discuss some of my recent work in both machine learning and neuroscience. Animal brains are some of the most complex objects in the known universe with a long history of mathematical modeling. Experimental advances in neuroscience offer exciting new datasets to test these mathematical theories. At the same time, computer engineers use similar principles to design surprisingly capable artificially intelligent systems. My goal is to understand how network structure, i.e. the pattern of connections between neurons, translates into function for different brain areas. I will focus on a theory of artificial neural networks developed using the language of reproducing kernel Hilbert spaces, which have well-understood mathematical and statistical properties. These kernel theories explain why network architectures may be well-suited to certain tasks. We are working to extend such theories to more complicated neural systems with feature learning and dynamics.

Everyone is welcome!

Refreshments will be served at 3:30 in Bond Hall 300