Melola Discovery

← Melola Discovery

Akses terbuka Jurnal SINTA 1

Analyzing the Effects of Combining Gradient Conflict Mitigation Methods in Multi-Task Learning

CommIT (Communication and Information Technology) Journal · Vol. 18 No. 1 (2024) · hlm. 99–107

DOI
10.21512/commit.v18i1.8905
ISSN
2460-7010, 1979-2484
Terbit
2024-04-29
Penerbit
Bina Nusantara University
Bahasa
eng
Lisensi
creativecommons.org/licenses/by-sa/4.0

Abstrak

Multi-task machine learning approaches involve training a single model on multiple tasks at once to increase performance and efficiency over multiple singletask models trained individually on each task. When such a multi-task model is trained to perform multiple unrelated tasks, performance can degrade significantly since unrelated tasks often have gradients that vary widely in direction. These conflicting gradients may destructively interfere with each other, causing weights learned during the training of some tasks to become unlearned during the training of others. The research selects three existing methods to mitigate this problem: Project Conflicting Gradients (PCGrad), Modulation Module, and Language-Specific Subnetworks (LaSS). It explores how the application of different combinations of these methods affects the performance of a convolutional neural network on a multi-task image classification problem. The image classification problem used as a benchmark utilizes a dataset of 4,503 leaf images to create two separate tasks: the classification of plants and the detection of disease from leaf images. Experiment results on this problem show performance benefits over singular mitigation methods, with a combination of PCGrad and LaSS obtaining a task-averaged F1 score of 0.84686. This combination outperforms individual mitigation approaches by 0.01870, 0.02682, and 0.02434 for PCGrad, Modulation Module, and LaSS, respectively in terms of F1 score.

Kata kunci

Jaringan kolaborasi para penulis

18 penulis · 33 tautan kepenulisan bersama

Arahkan kursor atau ketuk sebuah nama untuk menyorot rekan penulisnya. Klik dua kali untuk membuka halaman penulis.

  • Binus University
  • Organisasi berbeda
  • Organisasi sama
  • Afiliasi tidak diketahui
  • Garis lebih tebal = lebih banyak artikel bersama · lingkaran lebih besar = lebih banyak artikel
Tampilkan sebagai tabel
Rekan penulisOrganisasiArtikel bersama
Daniel D. AlbestaBinus University1
Michael JonathanBinus University1
Muhammad JawadBinus University1
Oktovianus HardiawanBinus University1
Kevin AlexanderBinus University1
Rayhan Ardiya DwantaraBinus University1
Raihan Muhammad NaufalBinus University1
Rilo Chandra PradanaBinus University1
Angel KarentiaBinus University1
Ryan AdipradanaBinus University1
Bagas Pradipabista NayogaBinus University1
Ryan SuryadiBinus University1
Heinrich Reagan SalimBinus University1
Chintya DeBinus University1
Nicholas Daniel PratamaputraBinus University1

Afiliasi berasal dari OpenAlex; metadata OAI-PMH jurnal tidak memuatnya. Penulis tanpa ID OpenAlex dicocokkan berdasarkan nama, sehingga dua orang bisa tergabung atau satu orang bisa terpisah.

Sumber metadata: OAI-PMH (metadata jurnal) https://journal.binus.ac.id/index.php/commit/oai, pertama di-harvest 8 Nov 2026, terakhir terlihat 16 Nov 2026.
Metadata asli (oai_dc)
<oai_dc:dc xmlns="http://www.openarchives.org/OAI/2.0/" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:oai_dc="http://www.openarchives.org/OAI/2.0/oai_dc/" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://www.openarchives.org/OAI/2.0/oai_dc/  http://www.openarchives.org/OAI/2.0/oai_dc.xsd">
	<dc:title xml:lang="en-US">Analyzing the Effects of Combining Gradient Conflict Mitigation Methods in Multi-Task Learning</dc:title>
	<dc:creator>Alison, Richard</dc:creator>
	<dc:creator>Jonathan, Welly</dc:creator>
	<dc:creator>Suhartono, Derwin</dc:creator>
	<dc:subject xml:lang="en-US">Gradient Conflict Mitigation Methods</dc:subject>
	<dc:subject xml:lang="en-US">Multi-Task Learning</dc:subject>
	<dc:subject xml:lang="en-US">Project Conflicting Gradients (PCGrad)</dc:subject>
	<dc:subject xml:lang="en-US">Modulation Module</dc:subject>
	<dc:subject xml:lang="en-US">Language-Specific Subnetworks (LaSS)</dc:subject>
	<dc:description xml:lang="en-US">Multi-task machine learning approaches involve training a single model on multiple tasks at once to increase performance and efficiency over multiple singletask models trained individually on each task. When such a multi-task model is trained to perform multiple unrelated tasks, performance can degrade significantly since unrelated tasks often have gradients that vary widely in direction. These conflicting gradients may destructively interfere with each other, causing weights learned during the training of some tasks to become unlearned during the training of others. The research selects three existing methods to mitigate this problem: Project Conflicting Gradients (PCGrad), Modulation Module, and Language-Specific Subnetworks (LaSS). It explores how the application of different combinations of these methods affects the performance of a convolutional neural network on a multi-task image classification problem. The image classification problem used as a benchmark utilizes a dataset of 4,503 leaf images to create two separate tasks: the classification of plants and the detection of disease from leaf images. Experiment results on this problem show performance benefits over singular mitigation methods, with a combination of PCGrad and LaSS obtaining a task-averaged F1 score of 0.84686. This combination outperforms individual mitigation approaches by 0.01870, 0.02682, and 0.02434 for PCGrad, Modulation Module, and LaSS, respectively in terms of F1 score.</dc:description>
	<dc:publisher xml:lang="en-US">Bina Nusantara University</dc:publisher>
	<dc:date>2024-04-29</dc:date>
	<dc:type>info:eu-repo/semantics/article</dc:type>
	<dc:type>info:eu-repo/semantics/publishedVersion</dc:type>
	<dc:type xml:lang="en-US">Peer-reviewed Article</dc:type>
	<dc:format>application/pdf</dc:format>
	<dc:identifier>https://journal.binus.ac.id/index.php/commit/article/view/8905</dc:identifier>
	<dc:identifier>10.21512/commit.v18i1.8905</dc:identifier>
	<dc:source xml:lang="en-US">CommIT (Communication and Information Technology) Journal; Vol. 18 No. 1 (2024): CommIT Journal ; 99-107</dc:source>
	<dc:source>2460-7010</dc:source>
	<dc:source>1979-2484</dc:source>
	<dc:language>eng</dc:language>
	<dc:relation>https://journal.binus.ac.id/index.php/commit/article/view/8905/5066</dc:relation>
	<dc:rights xml:lang="en-US">Copyright (c) 2024 Richard Alison, Welly Jonathan, Derwin Suhartono</dc:rights>
	<dc:rights xml:lang="en-US">https://creativecommons.org/licenses/by-sa/4.0</dc:rights>
</oai_dc:dc>