Talk Title
The mixing Time of Metropolized Hamiltonian Monte Carlo
Abstract
We analyze the mixing time of Metropolized Hamiltonian Monte Carlo (HMC) with the leapfrog integrator to sample from a distribution whose log-density is smooth, has Lipschitz Hessian in Frobenius norm and satisfies isoperimetry. We bound the gradient complexity to reach ϵ error in total variation distance from a warm start by Õ (d^{1/4} polylog(1/ϵ)) and demonstrate the benefit of choosing the number of leapfrog steps to be larger than 1.
Bio
Yuansi Chen is an assistant professor in the Department of Statistical Science at Duke University since Spring 2021. Previously, he was a postdoc fellow at ETH Foundations of Data Science (ETH-FDS) in ETH. He obtained his PhD in the Department of Statistics at UC Berkeley in 2019. He obtained his undergraduate degree from Ecole Polytechnique in France.
Yuansi’s main research interests include Markov chain Monte Carlo algorithms and their convergence guarantees in high dimensional statistical models, high dimensional convex geometry, domain adaptation and statistical challenges that arise in computational neuroscience.