INHOMOGENEOUS RANDOM SYSTEMS 26-27 January 2027 Institut Henri Poincaré 11-13, rue Pierre et Marie Curie, Paris

The aim of this annual workshop is to bring together mathematicians and physicists working on disordered or random systems, and to discuss recent developments on themes of common interest. Each day is devoted to a specific topic.
Note that the 46th edition of the "Journées de Physique Statistique" is scheduled on January 28th and 29th.

Tuesday 26 January: Beyond Detailed Balance: Nonreversible Markov Sampling.
        Moderator: Cristina Toninelli (Paris)

Over the last two decades, Monte Carlo methodology has undergone a striking transformation, driven by the convergence of ideas originating in several communities: computational statistical physics, through irreversible Monte Carlo and event-chain algorithms; probability theory, through the analysis of non-self-adjoint Markov semigroups and hypocoercive methods for the quantitative convergence of piecewise- deterministic processes; computational statistics, through the development of nonreversible tempering methods. The common starting point is the observation that detailed balance is sufficient, but not necessary, for a Markov process to sample a prescribed distribution: correctness requires only invariance. Once reversibility is abandoned, stationary probability currents may circulate through the state space. When appropriately designed, these currents suppress diffusive backtracking and transport the process more efficiently through the target distribution. The workshop aims to bring probability, numerical analysis, computational statistics, and statistical physics together around a common set of questions: how should nonreversible currents be designed; when can irreversibility change the scaling of relaxation rather than merely its prefactor; and how can these gains be proved and implemented efficiently?

Speakers: Sebastiano Grazzi (Milano), Werner Krauth (Paris), Francis Lörler (Leipzig), Brune Massoulié (Augsburg & München), Manon Michel (Clermont-Ferrand), Pierre Monmarché (Marne-la-Vallée).

There are longstanding connections between disordered magnetism and information theory, including connections with Bayesian inference, learning, and error correction. For example, the phase diagram of the random bond Ising model is relevant to tasks in image reconstruction, and to the stability of the toric code quantum memory, while there are deep connections between spin glasses and classical error-correcting codes. This meeting will explore synergies between inference, decoding, and disordered magnetism that have emerged more recently. Recent work has shed light on the physics of low-density parity check codes, while ideas from mean-field spin glasses have led to rapid progress in understanding high-dimensional inference problems. In another arena, the task of imaging or monitoring a many-body system, in order to extract data about the state, has been found to lead to nontrivial inference phase transitions. Rigorous work is also underway to characterize the difficulty of learning the properties of a quantum system from data. The workshop aims to discuss recent progress and encourage dialog between researchers from related areas.

Speakers: Ewan McCulloch (Paris), Tibor Rakovszky (Budapest), Cambyse Rouzé^(*) (Palaiseau), Guilhem Semerjian (Paris), Grace Sommers (New Haven), Romain Vasseur (Genève).
^(*) To be confirmed

Christian Maes	Kirone Mallick	Ellen Saada
KU Leuven	CEA, Saclay	CNRS & Université Paris Cité
christian.maes@kuleuven.be	kirone.mallick@ipht.fr	ellen.saada@mi.parisdescartes.fr

Program of previous sessions: 1994 - 1995 - 1996 - 1997 - 1998 (Moving fronts, Griffiths singularities) - 1999
2000 - 2001 - 2002 - 2003 - 2004 - 2005 - 2006 - 2007 - 2008 - 2009 - 2010 - 2011 - 2012 - 2013 - 2014 - 2015 - 2016 - 2017 - 2018 - 2019 - 2020 - 2021 - 2022 - 2023 - 2024 - 2025 - 2026