Fine-tuning can cripple your foundation model; preserving features may be the solution

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Mukhoti, Jishnu, Gal, Yarin, Torr, Philip H. S., Dokania, Puneet K.
Natura: Preprint
Pubblicazione: 2023
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866909235820888064
author Mukhoti, Jishnu
Gal, Yarin
Torr, Philip H. S.
Dokania, Puneet K.
author_facet Mukhoti, Jishnu
Gal, Yarin
Torr, Philip H. S.
Dokania, Puneet K.
contents Pre-trained foundation models, due to their enormous capacity and exposure to vast amounts of data during pre-training, are known to have learned plenty of real-world concepts. An important step in making these pre-trained models effective on downstream tasks is to fine-tune them on related datasets. While various fine-tuning methods have been devised and have been shown to be highly effective, we observe that a fine-tuned model's ability to recognize concepts on tasks $\textit{different}$ from the downstream one is reduced significantly compared to its pre-trained counterpart. This is an undesirable effect of fine-tuning as a substantial amount of resources was used to learn these pre-trained concepts in the first place. We call this phenomenon ''concept forgetting'' and via experiments show that most end-to-end fine-tuning approaches suffer heavily from this side effect. To this end, we propose a simple fix to this problem by designing a new fine-tuning method called $\textit{LDIFS}$ (short for $\ell_2$ distance in feature space) that, while learning new concepts related to the downstream task, allows a model to preserve its pre-trained knowledge as well. Through extensive experiments on 10 fine-tuning tasks we show that $\textit{LDIFS}$ significantly reduces concept forgetting. Additionally, we show that LDIFS is highly effective in performing continual fine-tuning on a sequence of tasks as well, in comparison with both fine-tuning as well as continual learning baselines.
format Preprint
id arxiv_https___arxiv_org_abs_2308_13320
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Fine-tuning can cripple your foundation model; preserving features may be the solution
Mukhoti, Jishnu
Gal, Yarin
Torr, Philip H. S.
Dokania, Puneet K.
Machine Learning
Computer Vision and Pattern Recognition
Pre-trained foundation models, due to their enormous capacity and exposure to vast amounts of data during pre-training, are known to have learned plenty of real-world concepts. An important step in making these pre-trained models effective on downstream tasks is to fine-tune them on related datasets. While various fine-tuning methods have been devised and have been shown to be highly effective, we observe that a fine-tuned model's ability to recognize concepts on tasks $\textit{different}$ from the downstream one is reduced significantly compared to its pre-trained counterpart. This is an undesirable effect of fine-tuning as a substantial amount of resources was used to learn these pre-trained concepts in the first place. We call this phenomenon ''concept forgetting'' and via experiments show that most end-to-end fine-tuning approaches suffer heavily from this side effect. To this end, we propose a simple fix to this problem by designing a new fine-tuning method called $\textit{LDIFS}$ (short for $\ell_2$ distance in feature space) that, while learning new concepts related to the downstream task, allows a model to preserve its pre-trained knowledge as well. Through extensive experiments on 10 fine-tuning tasks we show that $\textit{LDIFS}$ significantly reduces concept forgetting. Additionally, we show that LDIFS is highly effective in performing continual fine-tuning on a sequence of tasks as well, in comparison with both fine-tuning as well as continual learning baselines.
title Fine-tuning can cripple your foundation model; preserving features may be the solution
topic Machine Learning
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2308.13320