LORS: Low-rank Residual Structure for Parameter-Efficient Network Stacking

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Li, Jialin, Nie, Qiang, Fu, Weifu, Lin, Yuhuan, Tao, Guangpin, Liu, Yong, Wang, Chengjie
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866916150098526208
author Li, Jialin
Nie, Qiang
Fu, Weifu
Lin, Yuhuan
Tao, Guangpin
Liu, Yong
Wang, Chengjie
author_facet Li, Jialin
Nie, Qiang
Fu, Weifu
Lin, Yuhuan
Tao, Guangpin
Liu, Yong
Wang, Chengjie
contents Deep learning models, particularly those based on transformers, often employ numerous stacked structures, which possess identical architectures and perform similar functions. While effective, this stacking paradigm leads to a substantial increase in the number of parameters, posing challenges for practical applications. In today's landscape of increasingly large models, stacking depth can even reach dozens, further exacerbating this issue. To mitigate this problem, we introduce LORS (LOw-rank Residual Structure). LORS allows stacked modules to share the majority of parameters, requiring a much smaller number of unique ones per module to match or even surpass the performance of using entirely distinct ones, thereby significantly reducing parameter usage. We validate our method by applying it to the stacked decoders of a query-based object detector, and conduct extensive experiments on the widely used MS COCO dataset. Experimental results demonstrate the effectiveness of our method, as even with a 70\% reduction in the parameters of the decoder, our method still enables the model to achieve comparable or
format Preprint
id arxiv_https___arxiv_org_abs_2403_04303
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle LORS: Low-rank Residual Structure for Parameter-Efficient Network Stacking
Li, Jialin
Nie, Qiang
Fu, Weifu
Lin, Yuhuan
Tao, Guangpin
Liu, Yong
Wang, Chengjie
Computer Vision and Pattern Recognition
Deep learning models, particularly those based on transformers, often employ numerous stacked structures, which possess identical architectures and perform similar functions. While effective, this stacking paradigm leads to a substantial increase in the number of parameters, posing challenges for practical applications. In today's landscape of increasingly large models, stacking depth can even reach dozens, further exacerbating this issue. To mitigate this problem, we introduce LORS (LOw-rank Residual Structure). LORS allows stacked modules to share the majority of parameters, requiring a much smaller number of unique ones per module to match or even surpass the performance of using entirely distinct ones, thereby significantly reducing parameter usage. We validate our method by applying it to the stacked decoders of a query-based object detector, and conduct extensive experiments on the widely used MS COCO dataset. Experimental results demonstrate the effectiveness of our method, as even with a 70\% reduction in the parameters of the decoder, our method still enables the model to achieve comparable or
title LORS: Low-rank Residual Structure for Parameter-Efficient Network Stacking
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2403.04303