Sum-product networks: A survey

Sánchez-Cauce, Raquel; Paris, Iago; Vegas, F.J.

doi:10.1109/tpami.2021.3061898

Cited by 13 publications

(10 citation statements)

References 46 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Sum-Product Networks (SPN, more generally called a probabilistic circuit) [15], [16] is a machine learning workload that can tractably model joint probabilities of random variables. SPNs are often used in combination with neural networks, for various reasoning and perceptual tasks like noisetolerant image recognition [17], robotic navigation [18], robust human-activity detection [19], etc.…”

Section: Sum-product Networkmentioning

confidence: 99%

GraphOpt: Constrained-Optimization-Based Parallelization of Irregular Graphs

Shah

Meert

Verhelst

2022

IEEE Trans. Parallel Distrib. Syst.

View full text Add to dashboard Cite

Sparse, irregular graphs show up in various applications like linear algebra, machine learning, engineering simulations, robotic control, etc. These graphs have a high degree of parallelism, but their execution on parallel threads of modern platforms remains challenging due to the irregular data dependencies. The execution performance can be improved by efficiently partitioning the graphs such that the communication and thread synchronization overheads are minimized without hurting the utilization of the threads. To achieve this, this paper proposes GRAPHOPT, a tool that models the graph parallelization as a constrained optimization problem and uses the open Google OR-Tools solver to find good partitions. Several scalability techniques are developed to handle large real-world graphs with millions of nodes and edges. Extensive experiments are performed on the graphs of sparse matrix triangular solves (linear algebra) and sum-product networks (machine learning), respectively, showing a mean speedup of 2.0× and 1.8× over previous state-of-the-art libraries, demonstrating the effectiveness of the constrained-optimization-based graph parallelization.

show abstract

Section: Sum-product Networkmentioning

confidence: 99%

GraphOpt: Constrained-Optimization-Based Parallelization of Irregular Graphs

Shah

Meert

Verhelst

2022

IEEE Trans. Parallel Distrib. Syst.

View full text Add to dashboard Cite

show abstract

“…The other workload that is used to demonstrate the generality of GRAPHOPT is Sum-Product Network (SPN) [18]. An SPN (more generally called a probabilistic circuit) is a machine learning workload that can tractably model joint probabilities of random variables.…”

Section: Sum-product Networkmentioning

confidence: 99%

GRAPHOPT: constrained-optimization-based parallelization of irregular graphs

Shah,

Meert,

Verhelst

2021

Preprint

View full text Add to dashboard Cite

Sparse, irregular graphs show up in various applications like linear algebra, machine learning, engineering simulations, robotic control, etc. These graphs have a high degree of parallelism, but their execution on parallel threads of modern platforms remains challenging due to the irregular data dependencies. The parallel execution performance can be improved by efficiently partitioning the graphs such that the communication and thread synchronization overheads are minimized without hurting the utilization of the threads. To achieve this, this paper proposes GRAPHOPT, a tool that models the graph parallelization as a constrained optimization problem and uses the open Google OR-Tools solver to find good partitions. Several scalability techniques are developed to handle large real-world graphs with millions of nodes and edges. Extensive experiments are performed on the graphs of sparse matrix triangular solves (linear algebra) and sum-product networks (machine learning). GRAPHOPT achieves an average speedup of 1.4× and 6.1× over the Intel Math Kernel Library and a commonly-used heuristic partitioning technique, respectively, demonstrating the effectiveness of the constrained optimization-based graph parallelization.

show abstract

“…In the normal setting, while the data is centrally available, the weights on the edges of the sum nodes are determined by the following equation (see [1,Equation (24)]):…”

Section: Privacy-preserving Parameter Learningmentioning

confidence: 99%

“…Contrary to graphical models such as Bayesian networks, SPNs allow efficient inferences such as computing marginal and posterior probabilities, most probable explanation (MPE), approximate maximum aposteriori (MAP), and approximate MAX. For a good and detailed introduction to SPNs and these types of inferences, the reader is referred to [1].…”

Section: Introductionmentioning

confidence: 99%

Fast Private Parameter Learning and Inference for Sum-Product Networks

Althaus,

Dousti,

Kramer

et al. 2021

Preprint

View full text Add to dashboard Cite

A sum-product network (SPN) is a graphical model that allows several types of inferences to be drawn efficiently. There are two types of learning for SPNs: Learning the architecture of the model, and learning the parameters. In this paper, we tackle the second problem: We show how to learn the weights for the sum nodes, assuming the architecture is fixed, and the data is horizontally partitioned between multiple parties. The computations will preserve the privacy of each participant. Furthermore, we will use secret sharing instead of (homomorphic) encryption, which allows fast computations and requires little computational resources. To this end, we use a novel integer division to compute approximate real divisions. We also show how simple and private evaluations can be performed using the learned SPN.

show abstract

Sum-product networks: A survey

Cited by 13 publications

References 46 publications

GraphOpt: Constrained-Optimization-Based Parallelization of Irregular Graphs

GraphOpt: Constrained-Optimization-Based Parallelization of Irregular Graphs

GRAPHOPT: constrained-optimization-based parallelization of irregular graphs

Fast Private Parameter Learning and Inference for Sum-Product Networks

Contact Info

Product

Resources

About