Werner Krauth

lifted TASEP, an integrable example of non-reversible Markov chains

In recent years, non-reversible Markov chains have been at the origin of powerful Monte Carlo algorithm that are liberated from the detailed-balance condition. In this Oberseminar, I will discuss the lifted (1) TASEP (2), a Markov chain which is at the same time close to practical applications and integrable by the Bethe ansatz. The model describes one-dimensional particles on a lattice, with hard-sphere interactions. Its astonishing properties, which are only partially understood, are reminiscent of the ECMC (3) algorithms in higher dimensions.

(1) Lifted: as defined by Chen, Lovász and Pak (1999)
(2) TASEP: “Totally asymmetric simple exclusion process”
(3) ECMC: “event-chain Monte Carlo”