Francis Durand

PhD student in combinatorics · LAGA, Université Sorbonne Paris Nord · advised by Olivier Bodini

About

I am a third-year PhD student in mathematics and theoretical computer science (2024–2027). I work on the exact uniform random generation of combinatorial structures, at the interface between analytic combinatorics, probability and algorithms.

My main goal is to sample large structures in a quasi-entropic way: using a number of random bits as close as possible to the information-theoretic minimum given by Shannon's entropy, while keeping an optimal running time. Recent work covers set partitions and their generalisations (Dobiński structures), increasing trees, simply generated trees and binary words with a fixed number of ones.

Publications

  1. JournalDobiński Structures and Samplersto appear with Olivier Bodini Journal version, forthcoming. Extended version of the IWOCA 2025 paper: a general framework for sampling structures obtained by repeated grafting (set partitions, (r,s)-Stirling structures, rook placements, partial injections…).
  2. AofA 2026Efficient Sampling of Increasing Trees with Nadja Azzouz, Olivier Bodini and Bernhard Gittenberger LIPIcs, AofA 2026, article 3. DOIHAL
  3. AofA 2026Asymptotic Analysis of Generating Functions Arising from Dynamic Graphs with Nadja Azzouz, Olivier Bodini and Bernhard Gittenberger LIPIcs, AofA 2026, article 30. DOIHAL
  4. GASCom 2026Entropic Generation of Binary Words with Olivier Bodini EPTCS 445, pp. 38–46. DOIarXiv
  5. IWOCA 2025Optimal Random Bit Complexity in Efficient Sampling of Set Partition-Like Structures with Olivier Bodini LNCS 15885, pp. 405–417, Springer. DOIHAL
  6. GASCom 2024Optimal Generation of Strictly Increasing Binary Trees and Beyond with Olivier Bodini and Philippe Marchal arXivDOI

Operations research

  1. ICSP 2025Relative Value Iteration for Infinite-Horizon SDDP: Application to Hydroelectric Problem with B. Freitas Paulo da Costa, M. Azéma and V. Leclère

Talks & Presentations

Teaching & Supervision

Funding

My research is supported by the ANR–FWF project PAnDAG (ANR-23-CE48-0014).

Friends & Colleagues