Restructuring Tractable Probabilistic Circuits

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Zhang, Honghua, Wang, Benjie, Arenas, Marcelo, Broeck, Guy Van den
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866912354155888640
author Zhang, Honghua
Wang, Benjie
Arenas, Marcelo
Broeck, Guy Van den
author_facet Zhang, Honghua
Wang, Benjie
Arenas, Marcelo
Broeck, Guy Van den
contents Probabilistic circuits (PCs) are a unifying representation for probabilistic models that support tractable inference. Numerous applications of PCs like controllable text generation depend on the ability to efficiently multiply two circuits. Existing multiplication algorithms require that the circuits respect the same structure, i.e. variable scopes decomposes according to the same vtree. In this work, we propose and study the task of restructuring structured(-decomposable) PCs, that is, transforming a structured PC such that it conforms to a target vtree. We propose a generic approach for this problem and show that it leads to novel polynomial-time algorithms for multiplying circuits respecting different vtrees, as well as a practical depth-reduction algorithm that preserves structured decomposibility. Our work opens up new avenues for tractable PC inference, suggesting the possibility of training with less restrictive PC structures while enabling efficient inference by changing their structures at inference time.
format Preprint
id arxiv_https___arxiv_org_abs_2411_12256
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Restructuring Tractable Probabilistic Circuits
Zhang, Honghua
Wang, Benjie
Arenas, Marcelo
Broeck, Guy Van den
Artificial Intelligence
Machine Learning
Probabilistic circuits (PCs) are a unifying representation for probabilistic models that support tractable inference. Numerous applications of PCs like controllable text generation depend on the ability to efficiently multiply two circuits. Existing multiplication algorithms require that the circuits respect the same structure, i.e. variable scopes decomposes according to the same vtree. In this work, we propose and study the task of restructuring structured(-decomposable) PCs, that is, transforming a structured PC such that it conforms to a target vtree. We propose a generic approach for this problem and show that it leads to novel polynomial-time algorithms for multiplying circuits respecting different vtrees, as well as a practical depth-reduction algorithm that preserves structured decomposibility. Our work opens up new avenues for tractable PC inference, suggesting the possibility of training with less restrictive PC structures while enabling efficient inference by changing their structures at inference time.
title Restructuring Tractable Probabilistic Circuits
topic Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2411.12256