<i>HEAP</i>: A Holistic Error Assessment Framework for Multiple Approximations Using Probabilistic Graphical Models
Approximate computing has been a good paradigm of energy-efficient accelerator design. Accurate and fast error estimation is critical for appropriate approximate techniques selection so that power saving (or performance improvement) can be maximized with acceptable output quality in approximate acce...
Main Author: | |
---|---|
Format: | Article |
Language: | English |
Published: |
MDPI AG
2020-02-01
|
Series: | Electronics |
Subjects: | |
Online Access: | https://www.mdpi.com/2079-9292/9/2/373 |
id |
doaj-44eacc4a58184f51bef334c403205054 |
---|---|
record_format |
Article |
spelling |
doaj-44eacc4a58184f51bef334c4032050542020-11-25T02:51:11ZengMDPI AGElectronics2079-92922020-02-019237310.3390/electronics9020373electronics9020373<i>HEAP</i>: A Holistic Error Assessment Framework for Multiple Approximations Using Probabilistic Graphical ModelsJiajia Jiao0College of Information Engineering, Shanghai Maritime University, Shanghai 201306, ChinaApproximate computing has been a good paradigm of energy-efficient accelerator design. Accurate and fast error estimation is critical for appropriate approximate techniques selection so that power saving (or performance improvement) can be maximized with acceptable output quality in approximate accelerators. In the paper, we propose <i>HEAP</i>, a Holistic Error assessment framework to characterize multiple Approximate techniques with Probabilistic graphical models (PGM) in a joint way. <i>HEAP</i> maps the problem of evaluating errors induced by different approximate techniques into a PGM issue, including: (1) A heterogeneous Bayesian network is represented by converting an application’s data flow graph, where various approximate options are {precise, approximate} two-state X*-type nodes, while input or operating variables are {precise, approximate, unacceptable} three-state X-type nodes. These two different kinds of nodes are separately used to configure the available approximate techniques and track the corresponding error propagation for guaranteed configurability; (2) node learning is accomplished via an approximate library, which consists of probability mass functions of multiple approximate techniques to fast calculate each node’s Conditional Probability Table by mechanistic modeling or empirical modeling; (3) exact inference provides the probability distribution of output quality at three levels of precise, approximate, and unacceptable. We do a complete case study of 3 × 3 Gaussian kernels with different approximate configurations to verify <i>HEAP</i>. The comprehensive results demonstrate that <i>HEAP</i> is helpful to explore design space for power-efficient approximate accelerators, with just 4.18% accuracy loss and 3.34 × 10<sup>5</sup> speedup on average over Mentor Carlo simulation.https://www.mdpi.com/2079-9292/9/2/373error assessmentapproximate computingprobabilistic graphical modelsmultiple approximations |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Jiajia Jiao |
spellingShingle |
Jiajia Jiao <i>HEAP</i>: A Holistic Error Assessment Framework for Multiple Approximations Using Probabilistic Graphical Models Electronics error assessment approximate computing probabilistic graphical models multiple approximations |
author_facet |
Jiajia Jiao |
author_sort |
Jiajia Jiao |
title |
<i>HEAP</i>: A Holistic Error Assessment Framework for Multiple Approximations Using Probabilistic Graphical Models |
title_short |
<i>HEAP</i>: A Holistic Error Assessment Framework for Multiple Approximations Using Probabilistic Graphical Models |
title_full |
<i>HEAP</i>: A Holistic Error Assessment Framework for Multiple Approximations Using Probabilistic Graphical Models |
title_fullStr |
<i>HEAP</i>: A Holistic Error Assessment Framework for Multiple Approximations Using Probabilistic Graphical Models |
title_full_unstemmed |
<i>HEAP</i>: A Holistic Error Assessment Framework for Multiple Approximations Using Probabilistic Graphical Models |
title_sort |
<i>heap</i>: a holistic error assessment framework for multiple approximations using probabilistic graphical models |
publisher |
MDPI AG |
series |
Electronics |
issn |
2079-9292 |
publishDate |
2020-02-01 |
description |
Approximate computing has been a good paradigm of energy-efficient accelerator design. Accurate and fast error estimation is critical for appropriate approximate techniques selection so that power saving (or performance improvement) can be maximized with acceptable output quality in approximate accelerators. In the paper, we propose <i>HEAP</i>, a Holistic Error assessment framework to characterize multiple Approximate techniques with Probabilistic graphical models (PGM) in a joint way. <i>HEAP</i> maps the problem of evaluating errors induced by different approximate techniques into a PGM issue, including: (1) A heterogeneous Bayesian network is represented by converting an application’s data flow graph, where various approximate options are {precise, approximate} two-state X*-type nodes, while input or operating variables are {precise, approximate, unacceptable} three-state X-type nodes. These two different kinds of nodes are separately used to configure the available approximate techniques and track the corresponding error propagation for guaranteed configurability; (2) node learning is accomplished via an approximate library, which consists of probability mass functions of multiple approximate techniques to fast calculate each node’s Conditional Probability Table by mechanistic modeling or empirical modeling; (3) exact inference provides the probability distribution of output quality at three levels of precise, approximate, and unacceptable. We do a complete case study of 3 × 3 Gaussian kernels with different approximate configurations to verify <i>HEAP</i>. The comprehensive results demonstrate that <i>HEAP</i> is helpful to explore design space for power-efficient approximate accelerators, with just 4.18% accuracy loss and 3.34 × 10<sup>5</sup> speedup on average over Mentor Carlo simulation. |
topic |
error assessment approximate computing probabilistic graphical models multiple approximations |
url |
https://www.mdpi.com/2079-9292/9/2/373 |
work_keys_str_mv |
AT jiajiajiao iheapiaholisticerrorassessmentframeworkformultipleapproximationsusingprobabilisticgraphicalmodels |
_version_ |
1724735800179949568 |