DPP-PMRF: Rethinking Optimization for a Probabilistic Graphical Model Using Data-Parallel Primitives

2018/09/13 by Lessley, Brenton, Perciano, Talita, Heinemann, Colleen +3
#Distributed #FOS: Computer and information sciences #Parallel #and Cluster Computing (cs.DC)

paper · doi:10.48550/arxiv.1809.05018

Abstract

We present a new parallel algorithm for probabilistic graphical model optimization. The algorithm relies on data-parallel primitives (DPPs), which provide portable performance over hardware architecture. We evaluate results on CPUs and GPUs for an image segmentation problem. Compared to a serial baseline, we observe runtime speedups of up to 13X (CPU) and 44X (GPU). We also compare our performance to a reference, OpenMP-based algorithm, and find speedups of up to 7X (CPU).

Related

• Simple, Distributed, and Accelerated Probabilistic Programming
• Next-generation Probabilistic Computing Hardware with 3D MOSAICs, Illusion Scale-up, and Co-design
• Parallel Statistical Multi-resolution Estimation
• Speeding-up Graphical Model Optimization via a Coarse-to-fine Cascade of Pruning Classifiers
• MPLP++: Fast, Parallel Dual Block-Coordinate Ascent for Dense Graphical\n Models