Deep Generalized Convolutional Sum-Product Networks

Wolfshaar, Jos van de; Pronobis, Andrzej

doi:10.48550/arxiv.1902.06155

Cited by 1 publication

(3 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…LibSPN, 6 initiated by Andrzej Pronobis at the University of Washington, Seattle, WA [81], implements several methods for inference (marginal and conditional probabilities, and approximate MPE), parameter learning (batch and online, with GD and hard EM), and visualization of SPNs. It lacks algorithms for structural learning, but it allows building convolutional SPNs with a layer-oriented interface [82]. The SPNs, stored as Python structures, are compiled into Tensor-Flow graphs for parameter learning and inference; for this purpose LibSPN has implemented in C++ and CUDA some operations that cannot be performed efficiently with native TensorFlow operations.…”

Section: Software For Spnsmentioning

confidence: 99%

“…Several tutorials in Jupyter Notebook are available at its website. It has been used mainly for computer vision and robotics [67], [68], [82].…”

Section: Software For Spnsmentioning

confidence: 99%

“…Convolutional SPNs (ConvSPNs or CSPN) were developed independently in [82] and [93]. The first model exploits the inherent structure of spatial data in a similar way to convolutional neural networks by using the sum and product operations of SPNs.…”

Section: Extensions Of Spnsmentioning

confidence: 99%

See 2 more Smart Citations

Sum-product networks: A survey

Paris¹,

Sánchez-Cauce²,

Díez³

2020

Preprint

View full text Add to dashboard Cite

A sum-product network (SPN) is a probabilistic model, based on a rooted acyclic directed graph, in which terminal nodes represent univariate probability distributions and non-terminal nodes represent convex combinations (weighted sums) and products of probability functions. They are closely related to probabilistic graphical models, in particular to Bayesian networks with multiple context-specific independencies. Their main advantage is the possibility of building tractable models from data, i.e., models that can perform several inference tasks in time proportional to the number of links in the graph. They are somewhat similar to neural networks and can address the same kinds of problems, such as image processing and natural language understanding. This paper offers a survey of SPNs, including their definition, the main algorithms for inference and learning from data, the main applications, a brief review of software libraries, and a comparison with related models.

show abstract

Section: Software For Spnsmentioning

confidence: 99%

“…Several tutorials in Jupyter Notebook are available at its website. It has been used mainly for computer vision and robotics [67], [68], [82].…”

Section: Software For Spnsmentioning

confidence: 99%