Rates of convergence for a Gibbs sampler on the hypercube

In his work on partial exchangeability, de Finetti introduced a family of truncated Gaussian measures on \$[0,1]^d\$. One easy-to-implement procedure to get approximate samples from these measures is to run the Gibbs sampler, a Markov chain that suitably updates one (randomly chosen) coordinate at each step. In this talk, I will present some results and open questions concerning its mixing time. This is based on joint work with B. Gerencsér.