Characterizing the Energy Requirement of Computer Vision

The energy requirements of neural network learning are growing at a rapid rate. Increased energy demands have caused a global need to seek ways to improve energy efficiency of neural network learning. This thesis aims to establish a baseline on how adjusting basic parameters can affect energy consum...

Full description

Bibliographic Details
Main Author: Edelman, Daniel
Other Authors: Gadepally, Vijay N.
Format: Thesis
Published: Massachusetts Institute of Technology 2023
Online Access:https://hdl.handle.net/1721.1/151673
_version_ 1826213170186813440
author Edelman, Daniel
author2 Gadepally, Vijay N.
author_facet Gadepally, Vijay N.
Edelman, Daniel
author_sort Edelman, Daniel
collection MIT
description The energy requirements of neural network learning are growing at a rapid rate. Increased energy demands have caused a global need to seek ways to improve energy efficiency of neural network learning. This thesis aims to establish a baseline on how adjusting basic parameters can affect energy consumption in neural network learning on Computer Vision tasks. I catalogued the effects of various adjust adjustment from simple batch size adjustment to more complicated hardware configuration (such as power capping). Findings include that adjusting from single precision model to a mixed precision model can result in energy reductions of nearly 40%. Additionally power capping the GPU can reduce energy cost by an additional 10%.
first_indexed 2024-09-23T15:44:45Z
format Thesis
id mit-1721.1/151673
institution Massachusetts Institute of Technology
last_indexed 2024-09-23T15:44:45Z
publishDate 2023
publisher Massachusetts Institute of Technology
record_format dspace
spelling mit-1721.1/1516732023-08-01T03:09:46Z Characterizing the Energy Requirement of Computer Vision Edelman, Daniel Gadepally, Vijay N. Leiserson, Charles E. Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science The energy requirements of neural network learning are growing at a rapid rate. Increased energy demands have caused a global need to seek ways to improve energy efficiency of neural network learning. This thesis aims to establish a baseline on how adjusting basic parameters can affect energy consumption in neural network learning on Computer Vision tasks. I catalogued the effects of various adjust adjustment from simple batch size adjustment to more complicated hardware configuration (such as power capping). Findings include that adjusting from single precision model to a mixed precision model can result in energy reductions of nearly 40%. Additionally power capping the GPU can reduce energy cost by an additional 10%. M.Eng. 2023-07-31T19:58:01Z 2023-07-31T19:58:01Z 2023-06 2023-06-06T16:35:41.150Z Thesis https://hdl.handle.net/1721.1/151673 In Copyright - Educational Use Permitted Copyright retained by author(s) https://rightsstatements.org/page/InC-EDU/1.0/ application/pdf Massachusetts Institute of Technology
spellingShingle Edelman, Daniel
Characterizing the Energy Requirement of Computer Vision
title Characterizing the Energy Requirement of Computer Vision
title_full Characterizing the Energy Requirement of Computer Vision
title_fullStr Characterizing the Energy Requirement of Computer Vision
title_full_unstemmed Characterizing the Energy Requirement of Computer Vision
title_short Characterizing the Energy Requirement of Computer Vision
title_sort characterizing the energy requirement of computer vision
url https://hdl.handle.net/1721.1/151673
work_keys_str_mv AT edelmandaniel characterizingtheenergyrequirementofcomputervision