Résumé |
Markov-chain Monte Carlo is an outstanding computational tool in science. Since its origins, in 1953, it has
relied on the detailed-balance condition (that defines equilibrium) to map general computational problems
onto equilibrium-statistical-physics systems. Such reversible Markov chains are generally characterized by
diffusive transport. Non-reversible Markov chains violate the detailed-balance condition (they are by
definition out-of-equilibrium), and they are often characterized by ballistic transport. In recent years, the fast
approach of non-reversible Markov chains to equilibrium has fascinated scientists across several disciplines.
In this talk, I will introduce to this interdisciplinary field of research about non-equilibrium in equilibrium, and
discuss mathematical foundations, new algorithms, and exciting applications. In particular, I will discuss a
new class of one-dimensional Bethe-ansatz-solvable particle models as well as efficient (non-equilibrium)
algorithms that sample the Boltzmann distribution exp(-beta E) without ever evaluating the energy E.
|