Generating functions, which are widely used in combinatorics and probability theory, encode function values into the coefficients of a polynomial. In this paper, we explore their use as a tractable probabilistic model, and propose probabilistic generating circuits (PGCs) for their efficient representation. PGCs are strictly more expressive efficient than many existing tractable probabilistic models, including determinantal point processes (DPPs), probabilistic circuits (PCs) such as sum-product networks, and tractable graphical models. We contend that PGCs are not just a theoretical framework that unifies vastly different existing models, but also show great potential in modeling realistic data. We exhibit a simple class of PGCs that are not trivially subsumed by simple combinations of PCs and DPPs, and obtain competitive performance on a suite of density estimation benchmarks. We also highlight PGCs' connection to the theory of strongly Rayleigh distributions.

@inproceedings{ZhangICML21,
  author    = {Zhang, Honghua and Juba, Brendan and Van den Broeck, Guy},
  title     = {Probabilistic Generating Circuits},
  booktitle = {Proceedings of the 38th International Conference on Machine Learning (ICML)},
  month     = 7,
  year      = {2021},
  url       = "http://starai.cs.ucla.edu/papers/ZhangICML21.pdf",
  code = "https://github.com/joshuacnf/Probabilistic-Generating-Circuits",
  video = "https://slideslive.com/38958679/probabilistic-generating-circuits",
  annotation = "Long presentation, acceptance rate 166/5513 = 3\%",
  keywords  = {conference,selective}
}