Energy-efficient accelerator architecture for stereo image matching using approximate computing and statistical error compensation. Kim, E., P. & Shanbhag, N., R. In 2014 IEEE Global Conference on Signal and Information Processing, GlobalSIP 2014, 2014.
Paper abstract bibtex -Modern nanoscale processes exhibit stochastic be-havior that can no longer be ignored. Statistical error compen-sation (SEC) has shown significant benefits in achieving energy efficiency and error resiliency by embracing the stochastic nature of the underlying process. Approximate computing (AC), on the other hand, employs deterministic designs that produce imprecise results to achieve energy efficiency. In this paper, we bridge the two design paradigms by utilizing SEC and AC in the design of a machine learning accelerator core. ANT, a form of SEC, was applied to an AC based stereo image matching implementation in a 45 nm process. Simulation results show that ANT combined with AC achieves energy savings of 44% compared to a conventional system, and 32.7% compared to an AC only system, while its performance degradation is less than 4%. This result shows that embracing the stochasticity of the architecture is crucial in achieving high energy efficiency, and that AC and ANT are synergistic.
@inProceedings{
title = {Energy-efficient accelerator architecture for stereo image matching using approximate computing and statistical error compensation},
type = {inProceedings},
year = {2014},
identifiers = {[object Object]},
keywords = {Approximate computing,Low power,Statistical error compensation,Voltage overscaling},
id = {4ad4d3e9-c739-3036-9cee-12359a7df836},
created = {2016-07-27T00:10:58.000Z},
file_attached = {true},
profile_id = {f1db73c3-239d-36d4-8bb0-bab363f5c6ac},
group_id = {0a1f060f-2de4-3a9f-9ffc-8f142626b21a},
last_modified = {2016-07-27T00:10:58.000Z},
read = {false},
starred = {false},
authored = {false},
confirmed = {false},
hidden = {false},
abstract = {-Modern nanoscale processes exhibit stochastic be-havior that can no longer be ignored. Statistical error compen-sation (SEC) has shown significant benefits in achieving energy efficiency and error resiliency by embracing the stochastic nature of the underlying process. Approximate computing (AC), on the other hand, employs deterministic designs that produce imprecise results to achieve energy efficiency. In this paper, we bridge the two design paradigms by utilizing SEC and AC in the design of a machine learning accelerator core. ANT, a form of SEC, was applied to an AC based stereo image matching implementation in a 45 nm process. Simulation results show that ANT combined with AC achieves energy savings of 44% compared to a conventional system, and 32.7% compared to an AC only system, while its performance degradation is less than 4%. This result shows that embracing the stochasticity of the architecture is crucial in achieving high energy efficiency, and that AC and ANT are synergistic.},
bibtype = {inProceedings},
author = {Kim, Eric P. and Shanbhag, Naresh R.},
booktitle = {2014 IEEE Global Conference on Signal and Information Processing, GlobalSIP 2014}
}
Downloads: 0
{"_id":"auZGvSsg9yALTYgpd","bibbaseid":"kim-shanbhag-energyefficientacceleratorarchitectureforstereoimagematchingusingapproximatecomputingandstatisticalerrorcompensation-2014","downloads":0,"creationDate":"2016-07-27T00:03:58.623Z","title":"Energy-efficient accelerator architecture for stereo image matching using approximate computing and statistical error compensation","author_short":["Kim, E., P.","Shanbhag, N., R."],"year":2014,"bibtype":"inProceedings","biburl":null,"bibdata":{"title":"Energy-efficient accelerator architecture for stereo image matching using approximate computing and statistical error compensation","type":"inProceedings","year":"2014","identifiers":"[object Object]","keywords":"Approximate computing,Low power,Statistical error compensation,Voltage overscaling","id":"4ad4d3e9-c739-3036-9cee-12359a7df836","created":"2016-07-27T00:10:58.000Z","file_attached":"true","profile_id":"f1db73c3-239d-36d4-8bb0-bab363f5c6ac","group_id":"0a1f060f-2de4-3a9f-9ffc-8f142626b21a","last_modified":"2016-07-27T00:10:58.000Z","read":false,"starred":false,"authored":false,"confirmed":false,"hidden":false,"abstract":"-Modern nanoscale processes exhibit stochastic be-havior that can no longer be ignored. Statistical error compen-sation (SEC) has shown significant benefits in achieving energy efficiency and error resiliency by embracing the stochastic nature of the underlying process. Approximate computing (AC), on the other hand, employs deterministic designs that produce imprecise results to achieve energy efficiency. In this paper, we bridge the two design paradigms by utilizing SEC and AC in the design of a machine learning accelerator core. ANT, a form of SEC, was applied to an AC based stereo image matching implementation in a 45 nm process. Simulation results show that ANT combined with AC achieves energy savings of 44% compared to a conventional system, and 32.7% compared to an AC only system, while its performance degradation is less than 4%. This result shows that embracing the stochasticity of the architecture is crucial in achieving high energy efficiency, and that AC and ANT are synergistic.","bibtype":"inProceedings","author":"Kim, Eric P. and Shanbhag, Naresh R.","booktitle":"2014 IEEE Global Conference on Signal and Information Processing, GlobalSIP 2014","bibtex":"@inProceedings{\n title = {Energy-efficient accelerator architecture for stereo image matching using approximate computing and statistical error compensation},\n type = {inProceedings},\n year = {2014},\n identifiers = {[object Object]},\n keywords = {Approximate computing,Low power,Statistical error compensation,Voltage overscaling},\n id = {4ad4d3e9-c739-3036-9cee-12359a7df836},\n created = {2016-07-27T00:10:58.000Z},\n file_attached = {true},\n profile_id = {f1db73c3-239d-36d4-8bb0-bab363f5c6ac},\n group_id = {0a1f060f-2de4-3a9f-9ffc-8f142626b21a},\n last_modified = {2016-07-27T00:10:58.000Z},\n read = {false},\n starred = {false},\n authored = {false},\n confirmed = {false},\n hidden = {false},\n abstract = {-Modern nanoscale processes exhibit stochastic be-havior that can no longer be ignored. Statistical error compen-sation (SEC) has shown significant benefits in achieving energy efficiency and error resiliency by embracing the stochastic nature of the underlying process. Approximate computing (AC), on the other hand, employs deterministic designs that produce imprecise results to achieve energy efficiency. In this paper, we bridge the two design paradigms by utilizing SEC and AC in the design of a machine learning accelerator core. ANT, a form of SEC, was applied to an AC based stereo image matching implementation in a 45 nm process. Simulation results show that ANT combined with AC achieves energy savings of 44% compared to a conventional system, and 32.7% compared to an AC only system, while its performance degradation is less than 4%. This result shows that embracing the stochasticity of the architecture is crucial in achieving high energy efficiency, and that AC and ANT are synergistic.},\n bibtype = {inProceedings},\n author = {Kim, Eric P. and Shanbhag, Naresh R.},\n booktitle = {2014 IEEE Global Conference on Signal and Information Processing, GlobalSIP 2014}\n}","author_short":["Kim, E., P.","Shanbhag, N., R."],"urls":{"Paper":"http://bibbase.org/service/mendeley/f1db73c3-239d-36d4-8bb0-bab363f5c6ac/file/30843f21-42fe-5f03-23bd-48da029a678a/2014-Energy-efficient_accelerator_architecture_for_stereo_image_matching_using_approximate_computing_and_statistical_err.pdf.pdf"},"bibbaseid":"kim-shanbhag-energyefficientacceleratorarchitectureforstereoimagematchingusingapproximatecomputingandstatisticalerrorcompensation-2014","role":"author","keyword":["Approximate computing","Low power","Statistical error compensation","Voltage overscaling"],"downloads":0},"search_terms":["energy","efficient","accelerator","architecture","stereo","image","matching","using","approximate","computing","statistical","error","compensation","kim","shanbhag"],"keywords":["approximate computing","low power","statistical error compensation","voltage overscaling"],"authorIDs":[]}