INTERDISCIPLINARY STOCHASTIC PROCESSES COLLOQUIUM Tuesday February 17, room 60 Evans, 4.10 - 5.00pm Speaker: Persi Diaconis (Stanford University) Title: The fastest mixing Markov chain on a graph Abstract: Given a probability distribution on the vertices of a finite connected graph, choose edge weights so that the natural random walk has the largest possible spectral gap. By reformulating this problem as a problem in semidefinite programming, we can compute the solution in examples with one million edges. Numerics suggest conjectures, some of which we can prove. Joint work with Steve Boyd and Lin Xiao.