arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2304.01752cs.CVcs.CLcs.LG
收藏

面向视觉-语言模型的黑盒少样本适配

Black Box Few-Shot Adaptation for Vision-Language models

Samsung AI Cambridge(三星人工智能剑桥研究院);Queen Mary University of London(伦敦玛丽女王大学)

Yassine Ouali, Adrian Bulat, Brais Martinez, Georgios Tzimiropoulos

AI总结:

针对视觉-语言模型少样本适配需访问权重且计算昂贵的问题,提出黑盒线性特征对齐LFA,基于预计算特征闭式初始化并重排序优化,支持监督与无监督训练,实验表现超越软提示学习。

摘要:

通过对比学习训练以对齐视觉与语言模态的视觉-语言(V-L)模型,已被证明具有很强的少样本学习能力。软提示学习是少样本下游适配的首选方法,旨在弥合由新域引发的分布偏移所造成的模态鸿沟。尽管提示学习具有参数高效性,但它仍需访问模型权重,并且对于具有数十亿参数的大模型而言,在计算上可能不可行。为解决这些不足,本文描述了一种用于V-L少样本适配的黑盒方法,该方法:(a)在预先计算的图像和文本特征上运行,因此无需访问模型权重即可工作;(b)训练速度快数个数量级;(c)同时适用于有监督和无监督训练;(d)甚至可用于对齐由单模态模型计算得到的图像和文本特征。为实现这一点,我们提出了线性特征对齐(Linear Feature Alignment,LFA),这是一种用于目标域中V-L重新对齐的简单线性方法。LFA由一个最小二乘问题的闭式解初始化,随后通过最小化重排序损失进行迭代更新。尽管其方法简单,但我们在11个图像数据集和2个视频数据集上的大量实验表明,该方法甚至可以超越软提示学习方法。

英文摘要:

Vision-Language (V-L) models trained with contrastive learning to align the visual and language modalities have been shown to be strong few-shot learners. Soft prompt learning is the method of choice for few-shot downstream adaptation aiming to bridge the modality gap caused by the distribution shift induced by the new domain. While parameter-efficient, prompt learning still requires access to the model weights and can be computationally infeasible for large models with billions of parameters. To address these shortcomings, in this work, we describe a black-box method for V-L few-shot adaptation that (a) operates on pre-computed image and text features and hence works without access to the model's weights, (b) it is orders of magnitude faster at training time, (c) it is amenable to both supervised and unsupervised training, and (d) it can be even used to align image and text features computed from uni-modal models. To achieve this, we propose Linear Feature Alignment (LFA), a simple linear approach for V-L re-alignment in the target domain. LFA is initialized from a closed-form solution to a least-squares problem and then it is iteratively updated by minimizing a re-ranking loss. Despite its simplicity, our approach can even surpass soft-prompt learning methods as shown by extensive experiments on 11 image and 2 video datasets.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑