Geometric bounds for eigenvalues of reversible Markov chains
I will talk about how quickly does a system relax to thermal equibrium, based on a a paper by P Diaconis and D Stroock, Ann. Appl. Probab. 1 36-61. The central result involves a characterisation of bottlenecks in the graph of transition rates that defines the stochastic process. I will try to explain the result in elementary terms, show a couple of simple examples, and, if time permits (which it probably won't) sketch the proof.
This is a roughly weekly series of didactical blackboard talks focussing on some theoretical aspect of Condensed Matter, Biological, and Statistical Physics..