An Analysis of Current Continual Learning Algorithms in an Image Classification Context

Jess Wakelin; Noor Mohammedali

doi:10.1109/ISCSIC57216.2022.00019

An Analysis of Current Continual Learning Algorithms in an Image Classification Context

Jess Wakelin^*, Noor Mohammedali

^*Corresponding author for this work

Research output: Contribution to Book/Report › Conference Contribution › peer-review

Abstract

Artificial Intelligence aims to mimic natural intelligent learning by using lifelong-machine-learning, which allows an AI to train and learn over it's lifetime. Various algorithms have been suggested and developed to allow lifelong learning, these algorithms require deeper analysis, to evaluate and highlight performance benefits. In this research, we will study with three state-of-the-art algorithms for lifelong learning: Rehearsal, elastic-weight-consolidation and synaptic-intelligence. We do an analysis and evaluation of their performance in a multiple task experiment, using different amounts of data, measuring several performance metrics. We found that these algorithms are similar in performance, but some algorithms perform better than others with less data, or show good performance in task one, but not subsequent tasks. These algorithms could be built upon and improved in future research. The evaluation demonstrated in this research are in the image classification context.

Original language	English
Title of host publication	6th International Symposium on Computer Science and Intelligent Control
Publisher	IEEE Computer Society's Conference Publishing Services (CPS)
Pages	34-39
Number of pages	6
DOIs	https://doi.org/10.1109/ISCSIC57216.2022.00019
Publication status	Published - 14 Mar 2023
Event	The 6th International Symposium on Computer Science and Intelligent Control - Beijing, China Duration: 11 Nov 2022 → 13 Nov 2022 Conference number: 6 https://www.iscsic.org/

Conference

Conference	The 6th International Symposium on Computer Science and Intelligent Control
Abbreviated title	ISCSIC 2022
Country/Territory	China
City	Beijing
Period	11/11/22 → 13/11/22
Internet address	https://www.iscsic.org/

Bibliographical note

Publisher's notice for accepted manuscript: "© 2023 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works."

Keywords

Continuous Learning
Image Classification
Lifelong Machine Learning
Non-Continuous Learning
Machine Learning
Artificial Intelligence
Intelligent Agent

Access to Document

10.1109/ISCSIC57216.2022.00019

Wakelin_etal_IEEE_2022_An_Analysis_of_Current_Continual_Learning_Algorithms_in_an_Image_Classification_ContextAccepted author manuscript, 372 KBLicence: Unspecified

https://ieeexplore.ieee.org/document/10063552Licence: Unspecified

Cite this

@inproceedings{9336a547e31844398bfc0560c52546d0,

title = "An Analysis of Current Continual Learning Algorithms in an Image Classification Context",

abstract = "Artificial Intelligence aims to mimic natural intelligent learning by using lifelong-machine-learning, which allows an AI to train and learn over it's lifetime. Various algorithms have been suggested and developed to allow lifelong learning, these algorithms require deeper analysis, to evaluate and highlight performance benefits. In this research, we will study with three state-of-the-art algorithms for lifelong learning: Rehearsal, elastic-weight-consolidation and synaptic-intelligence. We do an analysis and evaluation of their performance in a multiple task experiment, using different amounts of data, measuring several performance metrics. We found that these algorithms are similar in performance, but some algorithms perform better than others with less data, or show good performance in task one, but not subsequent tasks. These algorithms could be built upon and improved in future research. The evaluation demonstrated in this research are in the image classification context.",

keywords = "Continuous Learning, Image Classification, Lifelong Machine Learning, Non-Continuous Learning, Machine Learning, Artificial Intelligence, Intelligent Agent",

author = "Jess Wakelin and Noor Mohammedali",

note = "Publisher's notice for accepted manuscript: {"}{\textcopyright} 2023 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.{"}; The 6th International Symposium on Computer Science and Intelligent Control , ISCSIC 2022 ; Conference date: 11-11-2022 Through 13-11-2022",

year = "2023",

month = mar,

day = "14",

doi = "10.1109/ISCSIC57216.2022.00019",

language = "English",

pages = "34--39",

booktitle = "6th International Symposium on Computer Science and Intelligent Control",

publisher = "IEEE Computer Society's Conference Publishing Services (CPS)",

url = "https://www.iscsic.org/",

}

Wakelin, J & Mohammedali, N 2023, An Analysis of Current Continual Learning Algorithms in an Image Classification Context. in 6th International Symposium on Computer Science and Intelligent Control. IEEE Computer Society's Conference Publishing Services (CPS), pp. 34-39, The 6th International Symposium on Computer Science and Intelligent Control , Beijing, China, 11/11/22. https://doi.org/10.1109/ISCSIC57216.2022.00019

TY - GEN

T1 - An Analysis of Current Continual Learning Algorithms in an Image Classification Context

AU - Wakelin, Jess

AU - Mohammedali, Noor

N1 - Conference code: 6

PY - 2023/3/14

Y1 - 2023/3/14

N2 - Artificial Intelligence aims to mimic natural intelligent learning by using lifelong-machine-learning, which allows an AI to train and learn over it's lifetime. Various algorithms have been suggested and developed to allow lifelong learning, these algorithms require deeper analysis, to evaluate and highlight performance benefits. In this research, we will study with three state-of-the-art algorithms for lifelong learning: Rehearsal, elastic-weight-consolidation and synaptic-intelligence. We do an analysis and evaluation of their performance in a multiple task experiment, using different amounts of data, measuring several performance metrics. We found that these algorithms are similar in performance, but some algorithms perform better than others with less data, or show good performance in task one, but not subsequent tasks. These algorithms could be built upon and improved in future research. The evaluation demonstrated in this research are in the image classification context.

AB - Artificial Intelligence aims to mimic natural intelligent learning by using lifelong-machine-learning, which allows an AI to train and learn over it's lifetime. Various algorithms have been suggested and developed to allow lifelong learning, these algorithms require deeper analysis, to evaluate and highlight performance benefits. In this research, we will study with three state-of-the-art algorithms for lifelong learning: Rehearsal, elastic-weight-consolidation and synaptic-intelligence. We do an analysis and evaluation of their performance in a multiple task experiment, using different amounts of data, measuring several performance metrics. We found that these algorithms are similar in performance, but some algorithms perform better than others with less data, or show good performance in task one, but not subsequent tasks. These algorithms could be built upon and improved in future research. The evaluation demonstrated in this research are in the image classification context.

KW - Continuous Learning

KW - Image Classification

KW - Lifelong Machine Learning

KW - Non-Continuous Learning

KW - Machine Learning

KW - Artificial Intelligence

KW - Intelligent Agent

U2 - 10.1109/ISCSIC57216.2022.00019

DO - 10.1109/ISCSIC57216.2022.00019

M3 - Conference Contribution

SP - 34

EP - 39

BT - 6th International Symposium on Computer Science and Intelligent Control

PB - IEEE Computer Society's Conference Publishing Services (CPS)

T2 - The 6th International Symposium on Computer Science and Intelligent Control

Y2 - 11 November 2022 through 13 November 2022

ER -

An Analysis of Current Continual Learning Algorithms in an Image Classification Context

Abstract

Conference

Bibliographical note

Keywords

Access to Document

Fingerprint

Cite this