Heat kernels in graphs: A journey from random walks to geometry, and back

 Heat kernels are one of the most fundamental concepts in physics and mathematics. In physics, the heat kernel is a fundamental solution of the heat equation and connects the Laplacian operator to the rate of heat dissipation. In spectral geometry, many fundamental techniques are based on heat kernels. In finite Markov chain theory, heat kernels correspond to continuous-time random walks and constitute one of the most powerful techniques in estimating the mixing time.

In this talk, we will briefly discuss this line of research and its relation to heat kernels in graphs. In particular, we will see how heat kernels are used to design the first nearly-linear time algorithm for finding clusters in real-world graphs. Some interesting open questions will be addressed in the end.