Salvato in:
Dettagli Bibliografici
Autori principali: Sangarya, Vishwesh, Bradford, Richard, Kim, Jung-Eun
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:https://arxiv.org/abs/2405.09600
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866910449042194432
author Sangarya, Vishwesh
Bradford, Richard
Kim, Jung-Eun
author_facet Sangarya, Vishwesh
Bradford, Richard
Kim, Jung-Eun
contents In this paper, we propose a predictive quantifier to estimate the retraining cost of a trained model in distribution shifts. The proposed Aggregated Representation Measure (ARM) quantifies the change in the model's representation from the old to new data distribution. It provides, before actually retraining the model, a single concise index of resources - epochs, energy, and carbon emissions - required for the retraining. This enables reuse of a model with a much lower cost than training a new model from scratch. The experimental results indicate that ARM reasonably predicts retraining costs for varying noise intensities and enables comparisons among multiple model architectures to determine the most cost-effective and sustainable option.
format Preprint
id arxiv_https___arxiv_org_abs_2405_09600
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Aggregate Representation Measure for Predictive Model Reusability
Sangarya, Vishwesh
Bradford, Richard
Kim, Jung-Eun
Machine Learning
Artificial Intelligence
Computer Vision and Pattern Recognition
Computers and Society
In this paper, we propose a predictive quantifier to estimate the retraining cost of a trained model in distribution shifts. The proposed Aggregated Representation Measure (ARM) quantifies the change in the model's representation from the old to new data distribution. It provides, before actually retraining the model, a single concise index of resources - epochs, energy, and carbon emissions - required for the retraining. This enables reuse of a model with a much lower cost than training a new model from scratch. The experimental results indicate that ARM reasonably predicts retraining costs for varying noise intensities and enables comparisons among multiple model architectures to determine the most cost-effective and sustainable option.
title Aggregate Representation Measure for Predictive Model Reusability
topic Machine Learning
Artificial Intelligence
Computer Vision and Pattern Recognition
Computers and Society
url https://arxiv.org/abs/2405.09600