MoIL: Momentum Imitation Learning for Efficient Vision-Language Adaptation

计算机科学 人工智能 适应(眼睛) 模仿 计算机视觉 机器学习 心理学 光学 物理 社会心理学
作者
Gen Luo,Yiyi Zhou,Minglang Huang,Tianhe Ren,Xiaoshuai Sun,Rongrong Ji
出处
期刊:IEEE Transactions on Pattern Analysis and Machine Intelligence [IEEE Computer Society]
卷期号:: 1-13 被引量:1
标识
DOI:10.1109/tpami.2024.3435790
摘要

Pre-training and fine-tuning have been the de-facto paradigm in vision-language domains. Along with the rapid growth of model sizes, fully fine-tuning these large-scale vision-language pre-training (VLP) models requires prohibitively expensive storage costs. To address this issue, recent advances in NLP offer a promising and efficient adaptation approach called LoRA, which aims to approximate the fine-tuning of large pre-trained model by updating low-rank parameters. Despite its effectiveness, we identify that LoRA suffers a large approximation error on VLP models and its optimization is also inefficient, which greatly limits its performance upper bound. In this paper, we mathematically prove that the approximation error of low-rank adaptation can be optimized by a new optimization objective, i.e., the weight distance between LoRA and fine-tuning. Based on this finding, we propose a novel PETL method for VLP models, namely momentum imitation learning (MoIL). Specifically, MoIL formulates PETL as a weight imitation learning process and directly optimize the approximation error bound of the low-rank adaptation. Based on this training scheme, we also explore a new hybrid approximation function to reduce the learning difficulty of low-rank adaptations. With these two novel designs, MoIL can greatly improve the optimization efficiency of the low-rank parameters on VLP models. We validate MoIL on three VLP models ranging from end-to-end network to two-stage network, and conduct extensive experiments on four VL tasks. Experimental results demonstrate superior performance and optimization efficiency of MoIL than existing PETL methods. For instance, by updating only 6.23% parameters, MoIL can even outperform full tuning by +2.3% on image-text matching task. Meanwhile, its inference efficiency and generalization ability is also validated by multiple VLP models, e.g., VLMO and VinVL.
最长约 10秒,即可获得该文献文件

科研通智能强力驱动
Strongly Powered by AbleSci AI
科研通是完全免费的文献互助平台,具备全网最快的应助速度,最高的求助完成率。 对每一个文献求助,科研通都将尽心尽力,给求助人一个满意的交代。
实时播报
1秒前
3秒前
3秒前
4秒前
4秒前
超级的如南完成签到,获得积分10
5秒前
6秒前
小星星发布了新的文献求助10
6秒前
6秒前
6秒前
7秒前
dudu发布了新的文献求助10
7秒前
9秒前
酷波er应助大白采纳,获得10
10秒前
10秒前
10秒前
12秒前
13秒前
科研通AI6.2应助文心采纳,获得10
13秒前
胖崽胖崽完成签到,获得积分10
16秒前
17秒前
17秒前
人间几月完成签到,获得积分20
17秒前
17秒前
科研通AI6.2应助yier采纳,获得10
18秒前
如是之人完成签到,获得积分10
18秒前
19秒前
19秒前
isabella发布了新的文献求助10
19秒前
19秒前
22秒前
lalalucky1发布了新的文献求助10
23秒前
23秒前
dudu完成签到,获得积分10
23秒前
聪聪冲冲发布了新的文献求助10
24秒前
新闻联播完成签到 ,获得积分10
25秒前
我爱学习发布了新的文献求助10
26秒前
欢呼的兰完成签到,获得积分10
26秒前
28秒前
28秒前
高分求助中
(应助此贴封号)【重要!!请各用户(尤其是新用户)详细阅读】【科研通的精品贴汇总】 10000
Effects of Two Weeks of Red Light Therapy on Choroidal Thickness and Axial Length in Young Adults 700
Positive Art Therapy Theory and Practice 600
Management and the Arts 510
Matrix Methods in Data Mining and Pattern Recognition Second Edition 510
Key mechanistic insights into the intramolecular C-H bond amination and double bond aziridination in sulfamate esters catalyzed by dirhodium tetracarboxylate complexes 500
The Neuroscience of Language 400
热门求助领域 (近24小时)
化学 材料科学 医学 生物 纳米技术 工程类 有机化学 化学工程 生物化学 计算机科学 内科学 物理 复合材料 催化作用 细胞生物学 无机化学 光电子学 物理化学 电极 基因
热门帖子
关注 科研通微信公众号,转发送积分 7671713
求助须知:如何正确求助?哪些是违规求助? 9238873
关于积分的说明 19897874
捐赠科研通 7241216
什么是DOI,文献DOI怎么找? 3285105
关于科研通互助平台的介绍 2443380
邀请新用户注册赠送积分活动 2287296