Academic Journal
View in EDS

MemOpt: a memory optimization method for deep learning model training based on dual intelligent reinforcement learning.

Bibliographic Details
Title: MemOpt: a memory optimization method for deep learning model training based on dual intelligent reinforcement learning.
Authors: Zeng, Yan1 (AUTHOR), Jiang, Changhui1 (AUTHOR), Huang, Chengchuang1 (AUTHOR), Zhu, Ying1 (AUTHOR), Lu, Xiaofei2 (AUTHOR), Yuan, Junfeng3 (AUTHOR), Zhang, Jilin3 (AUTHOR)
Source: Computer Journal. Jul2026, Vol. 69 Issue 7, p1109-1121. 13p.
Subject Terms: *Computer memory management, *Reinforcement learning, *Deep learning, *Artificial neural networks
Abstract: In recent years, with the rapid growth in the scale of datasets and neural network models, there has been a significant imbalance between the memory requirements during model training and the memory resources available on training devices. Existing memory optimization techniques like recomputation, memory swapping, and their adaptive combinations do not fully consider the structural information of the model and overlook the impact of application costs and timing on training efficiency. Addressing this issue, this paper proposes a memory optimization method called MemOpt, which uses dual-agent reinforcement learning to dynamically search for appropriate memory optimization strategies and execution timing based on model structure and device information. It can optimize memory without compromising accuracy while minimizing additional overhead. Experimental results show that the MemOpt method significantly increases the maximum batch size for model training by up to 8.7% and training throughput by up to 42.3% compared with baseline methods. In the future, this method may find better applications in large-scale neural networks. [ABSTRACT FROM AUTHOR]
Database: Academic Search Index
FullText Text:
  Availability: 0
CustomLinks:
  – Url: https://resolver.ebsco.com/c/fiv2js/result?sid=EBSCO:asx&genre=article&issn=00104620&ISBN=&volume=69&issue=7&date=20260701&spage=1109&pages=1109-1121&title=Computer Journal&atitle=MemOpt%3A%20a%20memory%20optimization%20method%20for%20deep%20learning%20model%20training%20based%20on%20dual%20intelligent%20reinforcement%20learning.&aulast=Zeng%2C%20Yan&id=DOI:10.1093/comjnl/bxag014
    Name: Full Text Finder (for New FTF UI) (ns324271)
    Category: fullText
    Text: Full Text Finder
    MouseOverText: Full Text Finder
Header DbId: asx
DbLabel: Academic Search Index
An: 196418936
RelevancyScore: 1431
AccessLevel: 6
PubType: Academic Journal
PubTypeId: academicJournal
PreciseRelevancyScore: 1430.76818847656
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: MemOpt: a memory optimization method for deep learning model training based on dual intelligent reinforcement learning.
– Name: Author
  Label: Authors
  Group: Au
  Data: <searchLink fieldCode="AR" term="%22Zeng%2C+Yan%22">Zeng, Yan</searchLink><relatesTo>1</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Jiang%2C+Changhui%22">Jiang, Changhui</searchLink><relatesTo>1</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Huang%2C+Chengchuang%22">Huang, Chengchuang</searchLink><relatesTo>1</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Zhu%2C+Ying%22">Zhu, Ying</searchLink><relatesTo>1</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Lu%2C+Xiaofei%22">Lu, Xiaofei</searchLink><relatesTo>2</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Yuan%2C+Junfeng%22">Yuan, Junfeng</searchLink><relatesTo>3</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Zhang%2C+Jilin%22">Zhang, Jilin</searchLink><relatesTo>3</relatesTo> (AUTHOR)
– Name: TitleSource
  Label: Source
  Group: Src
  Data: <searchLink fieldCode="JN" term="%22Computer+Journal%22">Computer Journal</searchLink>. Jul2026, Vol. 69 Issue 7, p1109-1121. 13p.
– Name: Subject
  Label: Subject Terms
  Group: Su
  Data: *<searchLink fieldCode="DE" term="%22Computer+memory+management%22">Computer memory management</searchLink><br />*<searchLink fieldCode="DE" term="%22Reinforcement+learning%22">Reinforcement learning</searchLink><br />*<searchLink fieldCode="DE" term="%22Deep+learning%22">Deep learning</searchLink><br />*<searchLink fieldCode="DE" term="%22Artificial+neural+networks%22">Artificial neural networks</searchLink>
– Name: Abstract
  Label: Abstract
  Group: Ab
  Data: In recent years, with the rapid growth in the scale of datasets and neural network models, there has been a significant imbalance between the memory requirements during model training and the memory resources available on training devices. Existing memory optimization techniques like recomputation, memory swapping, and their adaptive combinations do not fully consider the structural information of the model and overlook the impact of application costs and timing on training efficiency. Addressing this issue, this paper proposes a memory optimization method called MemOpt, which uses dual-agent reinforcement learning to dynamically search for appropriate memory optimization strategies and execution timing based on model structure and device information. It can optimize memory without compromising accuracy while minimizing additional overhead. Experimental results show that the MemOpt method significantly increases the maximum batch size for model training by up to 8.7% and training throughput by up to 42.3% compared with baseline methods. In the future, this method may find better applications in large-scale neural networks. [ABSTRACT FROM AUTHOR]
PLink https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=asx&AN=196418936
RecordInfo BibRecord:
  BibEntity:
    Identifiers:
      – Type: doi
        Value: 10.1093/comjnl/bxag014
    Languages:
      – Code: eng
        Text: English
    PhysicalDescription:
      Pagination:
        PageCount: 13
        StartPage: 1109
    Subjects:
      – SubjectFull: Computer memory management
        Type: general
      – SubjectFull: Reinforcement learning
        Type: general
      – SubjectFull: Deep learning
        Type: general
      – SubjectFull: Artificial neural networks
        Type: general
    Titles:
      – TitleFull: MemOpt: a memory optimization method for deep learning model training based on dual intelligent reinforcement learning.
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: Zeng, Yan
      – PersonEntity:
          Name:
            NameFull: Jiang, Changhui
      – PersonEntity:
          Name:
            NameFull: Huang, Chengchuang
      – PersonEntity:
          Name:
            NameFull: Zhu, Ying
      – PersonEntity:
          Name:
            NameFull: Lu, Xiaofei
      – PersonEntity:
          Name:
            NameFull: Yuan, Junfeng
      – PersonEntity:
          Name:
            NameFull: Zhang, Jilin
    IsPartOfRelationships:
      – BibEntity:
          Dates:
            – D: 01
              M: 07
              Text: Jul2026
              Type: published
              Y: 2026
          Identifiers:
            – Type: issn-print
              Value: 00104620
          Numbering:
            – Type: volume
              Value: 69
            – Type: issue
              Value: 7
          Titles:
            – TitleFull: Computer Journal
              Type: main
ResultId 1