For a path ,
by the Cauchy-Schwarz inequality. Average over , multiply by , and sum. Reversing the order of summation in the definition of the two Dirichlet forms gives
Let . The variational formula for variance gives
Take a nonconstant eigenfunction attaining the Rayleigh quotient for . Then
Choose the uniform distribution on shortest paths equivariantly under graph automorphisms. Automorphisms preserve distances and send uniform shortest paths to uniform shortest paths, so vertex transitivity makes
constant in . Summing this constant over vertices counts each path-edge incidence at most twice:
Thus , and each nonnegative summand satisfies
Compare simple random walk with the independent sampler , whose spectral gap is one. Both stationary distributions are uniform. Route each transition of the sampler along a uniformly selected shortest path. For a directed graph edge ,
Part i therefore bounds the congestion by
Part a, equivalently the Canonical paths comparison theorem, gives

Articles by others on the same topic (0)

There are currently no matching articles.