Our scientific output, published, accessible, contestable.

A company that claims a scientific core without producing regular scientific publications sits outside the established scientific regime. Publication is the historical mechanism through which a discipline constitutes itself, documents itself, opens itself to peer challenge, and is transmitted.

Imagine All The People operates within that regime. Our work on synthetic population generation by Maximum Entropy Relaxation, on calibration protocols, on extended NPORS benchmarks, on large-scale extensions through Persistent Contrastive Divergence: all of it is published in the open scientific archives (ArXiv), in the field's reference academic editions (Frontiers in AI), and at the domain's scientific conferences.

This page organizes access to that corpus. It references the recent and upcoming publications signed by our teams, the editorial contributions we carry in the field's major journals, and the historical work of our co-founder François Pachet, which forms the intellectual matrix of our discipline. Everything is accessible through direct links to the original publications: no mediation, no promotional summary, raw access to the scientific texts.

Published April 2026: Maximum Entropy Relaxation of Multi-Way Cardinality Constraints for Synthetic Population Generation. François Pachet and Jean-Daniel Zucker. ArXiv:2603.22558. Founding article formalizing the Maximum Entropy Relaxation approach for generating synthetic populations satisfying unit, binary and ternary multi-dimensional constraints. Introduction of the convex dual formulation, the L-BFGS solver, and the systematic comparison with generalized raking on NPORS-derived benchmarks with 4 to 40 attributes. → Read the article on ArXiv

In preparation: Balanced Exact Population Synthesis via Maximum Entropy Repair. François Pachet and Jean-Charles Régin. Methodological article in final preparation, combining Maximum Entropy Relaxation and exact constraint programming for generating synthetic populations under balanced constraints. Publication expected in 2026.

In preparation: Scalable Maximum Entropy Population Synthesis via Persistent Contrastive Divergence. Extension of the Pachet-Zucker framework to the regime of more than 20 attributes, where the exact computation of expectations becomes intractable, through stochastic estimation via Gibbs sampling. Publication expected in 2026, extending the NPORS benchmarks to a significantly broader class of real configurations.

Each of these publications, as soon as it is available, will be referenced here with direct access to the original articles on ArXiv or in the corresponding journals.

Beyond publishing articles, our team contributes to the editorial structuring of the emerging field of synthetic populations in the international scientific literature.

Co-editing the Research Topic Synthetic Populations in Frontiers in Artificial Intelligence. François Pachet co-edits, with Frédéric Kaplan (École Polytechnique Fédérale de Lausanne) and Mirko Degli Esposti (University of Bologna), the Research Topic Synthetic Populations published in Frontiers in Artificial Intelligence. This edition gathers the most significant international work on the generation, use, validation and ethical stakes of synthetic populations. To date, it is the most structured editorial contribution to the emergence of this scientific discipline. → See the Research Topic in Frontiers in AI

This co-editorship embodies our position at the heart of the field's international scientific community: we do not simply publish our own work, we help structure the scientific field itself, in dialogue with leading researchers from EPFL and the University of Bologna.