FlexGibbs: Reconfigurable Parallel Gibbs Sampling Accelerator for Structured Graphs
FlexGibbs: Reconfigurable Parallel Gibbs Sampling Accelerator for Structured Graphs
复制标题
FlexGibbs：用于结构化图的可重构并行 Gibbs 采样加速器
DOI：
--
复制
发表时间：
2019
期刊：
影响因子：
--
通讯作者：
Gu
中科院分区：
文献类型：
--
作者：
Glenn G. Ko;Yuji Chai;Rob A. Rutenbar;D. Brooks;Gu
许多人认为深度学习成功的关键因素之一是它与现有加速器（主要是GPU）的兼容性。虽然GPU擅长处理深度学习中常见的线性代数内核，但它们并不是处理无监督学习方法（如贝叶斯模型和推理）的最佳架构。作为一个步骤，实现更好地理解概率模型的架构，吉布斯抽样，贝叶斯推理最常用的算法之一，研究了集中在并行收敛到目标分布和参数化组件。我们提出了FlexGibbs，一个可重构的并行吉布斯采样推理加速器的结构图。我们设计了一个最佳的解决马尔可夫随机场任务的架构，使用一系列的并行吉布斯采样器，使色调度。我们表明，对于声源分离应用，在Xilinx Zync CPU-FPGA SoC的FPGA结构上配置的FlexGibbs实现了1048倍的Gibbs采样推理加速，并且在ARM Cortex-A53上运行时能耗降低了99.85%。
Many consider one of the key components to the success of deep learning as its compatibility with existing accelerators, mainly GPU. While GPUs are great at handling linear algebra kernels commonly found in deep learning, they are not the optimal architecture for handling unsupervised learning methods such as Bayesian models and inference. As a step towards, achieving better understanding of architectures for probabilistic models, Gibbs sampling, one of the most commonly used algorithms for Bayesian inference, is studied with a focus on parallelism that converges to the target distribution and parameterized components. We propose FlexGibbs, a reconfigurable parallel Gibbs sampling inference accelerator for structured graphs. We designed an architecture optimal for solving Markov Random Field tasks using an array of parallel Gibbs samplers, enabled by chromatic scheduling. We show that for sound source separation application, FlexGibbs configured on the FPGA fabric of Xilinx Zync CPU-FPGA SoC achieved Gibbs sampling inference speedup of 1048x and 99.85% reduction in energy over running it on ARM Cortex-A53.