Scaling Probabilistic Circuits via Data Partitioning

Jonas Seng, Florian P. Busch, Pooja Prasad, Devendra S. Dhami, Martin Mundt, Kristian Kersting · 2025

Read Scaling Probabilistic Circuits via Data Partitioning by Jonas Seng, Florian P. Busch, Pooja Prasad, Devendra S. Dhami, Martin Mundt, Kristian Kersting (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-08141.txt.

Description

Jonas Seng and co-authors teach probability circuits across many machines by slicing the dataset again and again. Their 2025 paper scales a model family prized for exact, fast answers to probability questions. Even so, training such circuits on big real data stays heavy.

Recursive slicing of scattered data turns out to mirror the circuit's own structure. The result is a federated frame that trains circuits on split sites without pooling raw data. One frame covers side-by-side, stacked, and mixed splits of data, a first for the field. Trials on large datasets show faster training and true scale. The authors recast federated learning as density estimation over spread data.

Catalog record

CREATORJonas Seng, Florian P. Busch, Pooja Prasad, Devendra S. Dhami, Martin Mundt, Kristian Kersting
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE58 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-08141

Browse the full library · About the Tower of Babel