Statistical robustness of Markov chain Monte Carlo accelerators
- 17 April 2021
- conference paper
- conference paper
- Published by Association for Computing Machinery (ACM) in Proceedings of the 26th ACM International Conference on Architectural Support for Programming Languages and Operating Systems
Abstract
Statistical machine learning often uses probabilistic models and algorithms, such as Markov Chain Monte Carlo (MCMC), to solve a wide range of problems. Probabilistic computations, often considered too slow on conventional processors, can be accelerated with specialized hardware by exploiting parallelism and optimizing the design using various approximation techniques. Current methodologies for evaluating correctness of probabilistic accelerators are often incomplete, mostly focusing only on end-point result quality ("accuracy"). It is important for hardware designers and domain experts to look beyond end-point "accuracy" and be aware of how hardware optimizations impact statistical properties. This work takes a first step toward defining metrics and a methodology for quantitatively evaluating correctness of probabilistic accelerators. We propose three pillars of statistical robustness: 1) sampling quality, 2) convergence diagnostic, and 3) goodness of fit. We apply our framework to a representative MCMC accelerator and surface design issues that cannot be exposed using only application end-point result quality. We demonstrate the benefits of this framework to guide design space exploration in a case study showing that statistical robustness comparable to floating-point software can be achieved with limited precision, avoiding floating-point hardware overheads.Keywords
Funding Information
- Intel Corporation
- Semiconductor Research Corporation (2020-AH-2935)
- National Science Foundation (CNS-1616947)
This publication has 53 references indexed in Scilit:
- AcMC 2Published by Association for Computing Machinery (ACM) ,2019
- VIBNNPublished by Association for Computing Machinery (ACM) ,2018
- CACTI 7ACM Transactions on Architecture and Code Optimization, 2017
- Joint Lung CT Image Segmentation: A Hierarchical Bayesian ApproachPLOS ONE, 2016
- A Database and Evaluation Methodology for Optical FlowInternational Journal of Computer Vision, 2010
- Quality of random number generators significantly affects results of Monte Carlo simulations for organic and biological systemsJournal of Computational Chemistry, 2010
- A new hardware architecture for sampling the exponential distribution2008 Canadian Conference on Electrical and Computer Engineering, 2008
- Ultra-Efficient (Embedded) SOC Architectures based on Probabilistic CMOS (PCMOS) TechnologyPublished by Institute of Electrical and Electronics Engineers (IEEE) ,2006
- General Methods for Monitoring Convergence of Iterative SimulationsJournal of Computational and Graphical Statistics, 1998
- Stochastic stereo matching over scaleInternational Journal of Computer Vision, 1989