TY - GEN
T1 - Demystifying Bayesian Inference Workloads
AU - Wang, Yu Emma
AU - Zhu, Yuhao
AU - Ko, Glenn G.
AU - Reagen, Brandon
AU - Wei, Gu Yeon
AU - Brooks, David
N1 - Funding Information:
The authors would like to thank the anonymous reviewers for their thoughtful comments and constructive suggestions. The authors would like to thank Bob Adolf, Glenn Holloway, Svilen Kanev, Lifeng Nai, Margo Seltzer, and Cliff Young for their feedback. This work was supported in part by the Center for Applications Driving Architectures (ADA), one of six centers of JUMP, a Semiconductor Research Corporation program co-sponsored by DARPA. The work was also partially supported by NSF grant # CCF-1438983 and Intel.
Publisher Copyright:
© 2019 IEEE.
PY - 2019/4/22
Y1 - 2019/4/22
N2 - The recent surge of machine learning has motivated computer architects to focus intently on accelerating related workloads, especially in deep learning. Deep learning has been the pillar algorithm that has led the advancement of learning patterns from a vast amount of labeled data, or supervised learning. However, for unsupervised learning, Bayesian methods often work better than deep learning. Bayesian modeling and inference works well with unlabeled or limited data, can leverage informative priors, and has interpretable models. Despite being an important branch of machine learning, Bayesian inference generally has been overlooked by the architecture and systems communities. In this paper, we facilitate the study of Bayesian inference with the development of BayesSuite, a collection of seminal Bayesian inference workloads. We characterize the power and performance profiles of BayesSuite across a variety of current-generation processors and find significant diversity. Manually tuning and deploying Bayesian inference workloads requires deep understanding of the workload characteristics and hardware specifications. To address these challenges and provide high-performance, energy-efficient support for Bayesian inference, we introduce a scheduling and optimization mechanism that can be plugged into a system scheduler. We also propose a computation elision technique that further improves the performance and energy efficiency of the workloads by skipping computations that do not improve the quality of the inference. Our proposed techniques are able to increase Bayesian inference performance by 5.8 × on average over the naive assignment and execution of the workloads.
AB - The recent surge of machine learning has motivated computer architects to focus intently on accelerating related workloads, especially in deep learning. Deep learning has been the pillar algorithm that has led the advancement of learning patterns from a vast amount of labeled data, or supervised learning. However, for unsupervised learning, Bayesian methods often work better than deep learning. Bayesian modeling and inference works well with unlabeled or limited data, can leverage informative priors, and has interpretable models. Despite being an important branch of machine learning, Bayesian inference generally has been overlooked by the architecture and systems communities. In this paper, we facilitate the study of Bayesian inference with the development of BayesSuite, a collection of seminal Bayesian inference workloads. We characterize the power and performance profiles of BayesSuite across a variety of current-generation processors and find significant diversity. Manually tuning and deploying Bayesian inference workloads requires deep understanding of the workload characteristics and hardware specifications. To address these challenges and provide high-performance, energy-efficient support for Bayesian inference, we introduce a scheduling and optimization mechanism that can be plugged into a system scheduler. We also propose a computation elision technique that further improves the performance and energy efficiency of the workloads by skipping computations that do not improve the quality of the inference. Our proposed techniques are able to increase Bayesian inference performance by 5.8 × on average over the naive assignment and execution of the workloads.
KW - Bayesian inference
KW - Machine learning
KW - Workload characterization
UR - http://www.scopus.com/inward/record.url?scp=85065429870&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85065429870&partnerID=8YFLogxK
U2 - 10.1109/ISPASS.2019.00031
DO - 10.1109/ISPASS.2019.00031
M3 - Conference contribution
AN - SCOPUS:85065429870
T3 - Proceedings - 2019 IEEE International Symposium on Performance Analysis of Systems and Software, ISPASS 2019
SP - 177
EP - 189
BT - Proceedings - 2019 IEEE International Symposium on Performance Analysis of Systems and Software, ISPASS 2019
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2019 IEEE International Symposium on Performance Analysis of Systems and Software, ISPASS 2019
Y2 - 24 March 2019 through 26 March 2019
ER -