Tractable Representation Learning with Probabilistic Circuits

Steven Braun [email protected], Sahil Sidheekh [email protected], Antonio Vergari [email protected], Martin Mundt [email protected], Sriraam Natarajan [email protected], Kristian Kersting [email protected] · 2025

Read Tractable Representation Learning with Probabilistic Circuits by Steven Braun [email protected], Sahil Sidheekh [email protected], Antonio Vergari [email protected], Martin Mundt [email protected], Sriraam Natarajan [email protected], Kristian Kersting [email protected] (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-2507-04385.txt.

Description

Steven Braun and five co-authors bring representation learning to probabilistic circuits, a family of models prized because their inferences are exact and fast. Neural networks dominate the learning of compact data representations, called embeddings. Circuits have mostly borrowed embeddings from outside neural tools instead of learning their own.

Their framework, autoencoding probabilistic circuits, models the data and its embedding together inside the circuit. The embedding is then read out through the circuit's own exact inference. A neural decoder rebuilds the data from the embedding, and the whole system trains end to end. Because the circuit natively copes with missing values, the method proved sturdy when data had gaps. In tests it beat earlier circuit-based autoencoders on reconstruction quality. Its embeddings rivaled those of neural autoencoders, with better handling of missing data.

Catalog record

CREATORSteven Braun [email protected], Sahil Sidheekh [email protected], Antonio Vergari [email protected], Martin Mundt [email protected], Sriraam Natarajan [email protected], Kristian Kersting [email protected]
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE105 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2507-04385

Browse the full library · About the Tower of Babel