Salvato in:
Dettagli Bibliografici
Autori principali: Yang, Yi, Ikemura, Kei, Zhang, Qingwen, Zhu, Xiaomeng, Li, Ci, Batool, Nazre, Mansouri, Sina Sharif, Folkesson, John
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:https://arxiv.org/abs/2508.13979
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866911111226327040
author Yang, Yi
Ikemura, Kei
Zhang, Qingwen
Zhu, Xiaomeng
Li, Ci
Batool, Nazre
Mansouri, Sina Sharif
Folkesson, John
author_facet Yang, Yi
Ikemura, Kei
Zhang, Qingwen
Zhu, Xiaomeng
Li, Ci
Batool, Nazre
Mansouri, Sina Sharif
Folkesson, John
contents Recent multi-task learning studies suggest that linear scalarization, when using well-chosen fixed task weights, can achieve comparable to or even better performance than complex multi-task optimization (MTO) methods. It remains unclear why certain weights yield optimal performance and how to determine these weights without relying on exhaustive hyperparameter search. This paper establishes a direct connection between linear scalarization and MTO methods, revealing through extensive experiments that well-performing scalarization weights exhibit specific trends in key MTO metrics, such as high gradient magnitude similarity. Building on this insight, we introduce AutoScale, a simple yet effective two-phase framework that uses these MTO metrics to guide weight selection for linear scalarization, without expensive weight search. AutoScale consistently shows superior performance with high efficiency across diverse datasets including a new large-scale benchmark.
format Preprint
id arxiv_https___arxiv_org_abs_2508_13979
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle AutoScale: Linear Scalarization Guided by Multi-Task Optimization Metrics
Yang, Yi
Ikemura, Kei
Zhang, Qingwen
Zhu, Xiaomeng
Li, Ci
Batool, Nazre
Mansouri, Sina Sharif
Folkesson, John
Machine Learning
Recent multi-task learning studies suggest that linear scalarization, when using well-chosen fixed task weights, can achieve comparable to or even better performance than complex multi-task optimization (MTO) methods. It remains unclear why certain weights yield optimal performance and how to determine these weights without relying on exhaustive hyperparameter search. This paper establishes a direct connection between linear scalarization and MTO methods, revealing through extensive experiments that well-performing scalarization weights exhibit specific trends in key MTO metrics, such as high gradient magnitude similarity. Building on this insight, we introduce AutoScale, a simple yet effective two-phase framework that uses these MTO metrics to guide weight selection for linear scalarization, without expensive weight search. AutoScale consistently shows superior performance with high efficiency across diverse datasets including a new large-scale benchmark.
title AutoScale: Linear Scalarization Guided by Multi-Task Optimization Metrics
topic Machine Learning
url https://arxiv.org/abs/2508.13979